3.1.1 Purpose and Strategy of Efficiency Testing for Troubleshooting

The incentive for efficiency-testing a poorly performing column is obvious. The simulation with the expected efficiency shows how the tower should perform, while that based on the measured plant data provides the real efficiency. In the absence of flooding, a real efficiency below the expected one signifies issues such as maldistribution or channeling in trays and packings, poor hydraulic design, fouling, corrosion, or damage. In many cases (e.g., 195, 197, 234, 286, 351), a mismatch between simulation and plant data, or a low efficiency obtained from good plant data, was the key for correct diagnosis and an effective solution to an operating problem.

A well-documented performance/efficiency test report is essential even for a well-performing unit, for the following reasons:

1. This report provides an excellent starting point for almost every troubleshooting assignment, present or future, and highlights unit limitations and operating deficiencies.
   
2. When the tower malfunctions at a future date, availability of good operation test data dramatically reduces the troubleshooting effort.
   
3. A tower may appear to perform well, even while running at non-optimum conditions; for instance, energy usage may be excessive. Because of overdesign, these factors may be hidden. Yarborough et al. (515) presented several case histories (not all distillation-related) where plant testing directly led to major improvements in profitability.
   
4. Excess reflux and reboil due to poor efficiency may go unnoticed while generating excessive hydraulic loads and a premature capacity bottleneck. When unappreciated, an incorrect, or ineffective, or unnecessarily expensive debottleneck action may result.
   
5. Simulations used to determine the best running conditions and assess the effectiveness of the proposed modifications may be misleading unless tested against reliable data.
   
6. A discrepancy between the simulation and the tower performance (e.g., differences between the simulated and measured temperature profiles) may identify a hidden problem (351).
   
7. Basing a revamp on a reliable simulation, with efficiency tuned to correctly match the plant data, can be an enormous cost-saver. In two cases (277, 289), an expensive new tower was (or was to be) constructed based on a simulation that incorrectly reflected tower performance. In both cases, a correct simulation achieved (or would have achieved) the revamp by an inexpensive, minor modification. The author is familiar with many other similar cases.
   
8. A reliable efficiency determination provides a good basis for simulation and future design or optimization of similar towers and valuable input for the company database.
   
9. It provides a check that the contractor and/or internals vendor meets their guarantees.

Good performance/efficiency testing is rigorous and effort- and time-consuming. The cost-effectiveness of the rigorous procedure is often questioned, and a shortcut version is advocated. A suitable shortcut procedure can evolve from the rigorous procedure outlined here and in Refs. 4 and 192 by skipping over guidelines that are considered less critical.

The best procedure to adopt depends on the objective of the test. A shortcut test is best suited for detecting gross abnormalities and is often performed as a part of a troubleshooting effort. When investigating a gross malfunction, rigorous testing may slow down the identification and rectification of the fault. When a column appears to perform well, there is no urgency, so the focus should be on obtaining reliable data for future reference, favoring a rigorous test.

A shortcut test is unsuitable and often misleading for detecting subtle abnormalities, for obtaining column efficiency, for checking the design, for optimization, and as a basis for debottlenecking modifications. The author is familiar with many cases where shortcut tests applied for these purposes needed repeats, yielded conflicting data, provided inconclusive results, and led to ineffective modifications. In most of these cases, the test objectives were not met even though the total time, effort, and expense spent were severalfold those that would have been spent had a single rigorous test been performed. The author therefore strongly warns against applying a shortcut procedure for these purposes. This recommendation is shared by others (515).

Hanson, Golden, and Martin (126, 153) strongly warn against using shortcut tests, as well as using the plant management information system (MIS) or plant information (PI), as the sole basis for performing a revamp. They question the validity of some of these data, the ability of the statistical methods to detect faulty instruments, and the absence of key data not gathered by the plant systems. Sloley and Fraser (446) add concerns about MIS data gathering being often designed to analyze data trends, most suited for plant operation, rather than the absolute values needed for evaluation and revamps. Rigorous testing can fully take care of these concerns. “The surest way to have a successful, low investment revamp is to carefully (rigorously and correctly) define your current operating basis” (126). The author endorses their recommendation.

Shortcut tests range from those that do little more than taking a set of readings and samples from the column to those that incorporate checks of material, component and energy balances, and key instrumentation. Even for shortcut tests, the author recommends incorporating the above checks and spending time on preparations to ensure that key indicators provide adequate data. The key items can be extracted from the list of preparations and checks recommended here and in Refs. 4, 192.

3.1.2 Planning and Execution of Efficiency Testing for Troubleshooting

3.1.3 Preparations for Efficiency Testing

This is the most important phase of the test. An erroneous meter, a leaking block valve, or an incorrect lab analysis during the main test can dramatically reduce the reliability of the results, defeating the purpose of the test. This is the time to eliminate all potential problems and complications. The following considerations are important and are based on the author’s experience as well as many literature sources (4, 47, 121, 139, 153, 286, 319, 418, 426, 447, 465, 515).

1. Prepare a needs list for work required.
   

   This should include a detailed list of instruments to be checked and required laboratory analyses. Consider the following:

   
   
   
   
   • Thoroughly check for leaks. Closely study the P&ID and look at any lines connected to the tower, or cross-connecting products. Check that all the block valves on lines connecting different products with each other are shut. Use a temperature gun to make sure they are at ambient temperatures and not passing fluids. Look for any unmetered connections to the tower or its heat exchangers and make sure they are fully isolated and not leaking. See Section 3.2.1 for how unexpected leakages affect product purities.
     
   • Thoroughly check that all streams entering the tower are accounted for. Audit any purges entering or leaving. Devise ways of accounting for them or, if practical, have them blocked in for the test period. In one case (41), a small product stream was cooled, used as seal oil to the vacuum pump, and then returned to the column. The cooling of the stream had a significant impact on the column heat balance. In another case, a seemingly benign unmetered vapor equalizing line from a column overhead to the feed drum accounted for a 10–12% error in the tower mass balance (136).
     
   • Ensure that no important meters are left out and that the meters included are sufficient to entirely define unit performance. Add temporary instrumentation to fill any gaps. If in doubt, include the meter. It is much easier to ignore data that were collected but not needed than to generate needed data that were not collected (418). Include control valve openings as well as pump suction and discharge pressures that can be used to verify important flow rates. Seemingly unimportant information may turn out to be the deciding factor between contradicting measurements.
     
   • Closely review the literature on column instrumentation basics and pitfalls. Excellent coverage by Sands (410) and other resources (4, 192) should be closely studied, and any uncertainties that may adversely affect the validity and accuracy of key instrumentation used in the test should be addressed. Proper flow meter corrections should be applied. Ref. 361 has superb guidelines for and listing of flow corrections for a wide variety of flow meters.

     

     
     
   • Prepare a list of all handheld monitoring devices (e.g., timers, pyrometers, ultrasonic flow meters, electronic pressure gauges) required. Check their availability and have them ordered as necessary. Review any special requirements for using them. For instance, an ultrasonic flow meter should not be placed too close to an elbow as flow patterns created by the elbow may reduce reading accuracy. Unlike the installation in Figure 3.2, it was recommended (444) that it should be installed >15 pipe diameters from a single elbow, >40 pipe diameters downstream of a tee, >5 pipe diameters upstream of an elbow, and >10 diameters upstream of a tee. Confirm that the transducer is properly fixed and gelled to the pipe (435). Seek ways of verifying the measurement. A thorough testing program was shown to lead to high confidence in the ultrasonic meter readings (159). For split lines or manifolds, consider whether it is practical and desirable to measure the flows in each arm using an ultrasonic flow meter, as was demonstrated in one case (471).
     
   • Check whether any scaffolding is required to obtain the desired readings and have them ordered.
     
   • Ensure all meters are operational and read on-scale. Look for oddities in field instrument installation. A flow meter will read inaccurately when an orifice plate is incorrectly sized or damaged or when a tap is plugged. Ensure that all orifice plates are of the proper size; if practical, inspect them to ensure good condition. Failure to check orifice plates has ruined the results of many test runs (4, 192, 465). Audit meters on steam condensate lines; they may become erratic (e.g., 330) when flashing occurs in the condensate lines. All orifice taps should be blown. Inspect pressure indicator installations and check their taps for blockage. A pressure gauge located near an elbow or a tee will read the impact pressure rather than the fluid pressure. In Cases 25.1 and 25.3 in Ref. 201, poor calibration or installation of pressure transmitters misled troubleshooting. A vapor feed or reboiler return blowing into a pressure tap will add the velocity head to the pressure measurement; in one case (Case 4, Section 8.16), this made a difference of 5 psi. Inspect thermocouples and thermowells; they may read incorrectly if too short (as a thermowell described in Ref. 442 did) or in the presence of deposits. Check that all recorded temperatures for your test are actual, not pressure-compensated, temperatures. Inspect impulse lines, heat-tracing equipment, and vibrations near transmitters. Correct any oddities prior to the test.
     
   • Ensure all flow and differential pressure indicators are zeroed and calibrated. All dial thermometers and outside pressure gauges that can be safely removed with no leakage should be removed, and their accuracy checked. The unsatisfactory ones should be either replaced or removed before the test, and an alternative way of taking the measurement should be considered. Many unit thermowells are installed without a temperature-sensing element. These wells can be probed with a handheld device, or a temporary thermocouple can be installed in them to provide valuable data.
     
   • Whenever possible, verify flow rates by tank gauging. A reliable determination requires liquid density as well as correct height readings at the beginning, end, and additional time intervals. Identify additional sampling and temperature measurements needed to obtain liquid density, and check that you have accurate dimensions of the tanks.
     
   • Identify the location of sample points to be used in the test. Add cooling coils where required. Samples should be taken where a stream is either all liquid or all vapor and from a location where the stream is well mixed. Liquid samples normally give better accuracy except when flashing of lights is an issue. For sampling vapor lines, consider the possibility of atmospheric condensation, and for sampling liquid lines, consider that they may contain vapor bubbles.
     
   • Check whether samples can be safely drawn from the column; these can become very useful. For samples drawn from the column, phase separation may be required, because some vapor is likely to be present even in the bottom of the downcomer, while some liquid is likely to be present in the vapor space between the trays. Avoid drawing samples at the feed tray (465); such samples are likely to be dominated by the feed and are seldom representative of the tower composition. In packed towers, drawing samples from a collector right above the feed zone is usually good and informative.
     
   • Ensure all sampling lines are operable, free of blockage and comply with the relevant safety, health, and environmental requirements. Valuable sets of guidelines for sampling are available elsewhere (4, 319).
     
   • Prepare a list of all samples required for the test and their frequency. Together with the lab supervisors, determine the sample containers suitable for each analysis (e.g., bombs, glass bottles, plastic bottles) and the best way of sealing the containers (plugs, corks, screwed tops). Samples to be transported to outside labs may require special containers. Review the sampling procedure with the lab supervisor. The procedure should minimize flashing of lights from liquid samples. If vapor is sampled under pressure, the purge valve should be at the outlet to avoid Joule–Thompson condensation in the sample bottle. Determine which samples need special handling (e.g., refrigeration, stabilization, settling out of a second liquid phase). Attempt to use as few types of sample containers as possible to minimize errors. It has been recommended (447) that each sample taken should be large enough to run the analysis at least three times.
     
   • Check that the on-line analyzers are adequately calibrated and read the same as the lab analyzers. Review the records of these analyzers over the past three months. Straight lines and nonsensical numbers indicate a bad analyzer. Investigate the reason and ensure the problem was fixed and is unlikely to reoccur during the test. Check that the trends indicated by the analyzer make sense and that it properly responds to process changes.
     
   • Verify that the basis for the compositions reported by the lab analysis is valid and applicable throughout the expected composition range. Check that none of the reported compositions relies on extrapolation of data. This is addressed in detail by Kunz (270) for ethanol–water and methanol–water systems. For instance, inferring composition from liquid density measurements for ethanol–water mixtures above 60°C is based on extrapolation rather than data. Also, maxima occur in the refractive index–composition plots for these systems, so a measured refractive index near the maxima can indicate two different compositions.
     
   • Check that the range of component concentrations of the important components in all samples falls within the ranges of the analyzers available in the lab. A measurement of “< 0.1%” for an important component can make the entire test futile. Should any important components fall outside the measurable range, consider sending the samples to an outside lab or altering the test conditions to increase this concentration.
     
   • Heavy component groups are often reported as “C_x plus,” where x is some number between 5 and 12. Evaluate whether it is important to have an analysis of the split of these components. In one aromatic tower, it was found that assumptions regarding the split of the heavy non-key C₉+ components varied simulation results significantly, and analysis was necessary (278).
     
   • Refinery distillation product analyses are usually carried out either by chromatographic simulated distillation (ASTM D2887, D7169, and D3710) or by laboratory distillation (ASTM D86 and D1160). Occasionally, the laborious true boiling point (TBP) laboratory distillation (ASTM 2892) is used. Consider the pros and cons of each method for the test, and choose the preferred method for each tower. Ref. 154 provides some useful guidelines. Most commercial simulators use a TBP curve to define a series of individual pseudo-components. This is done by breaking down the TBP curve into small ranges so that the volumetric average temperature for each range corresponds to a defined TBP cut point. Simulations also calculate pseudo-component gravity from an API gravity curve, which can be obtained from lab analysis. If unavailable, the simulators use a less accurate approximation. Therefore, it is important to include the API gravity analyses (154).
     
   • In refinery gas plant towers as well as some petrochemical towers (debutanizers, depentanizers, naphtha splitters, aromatic towers), whenever practical, obtain full component analyses of the feeds and products (235, 278). As described in step 17 of Section 3.1.6, these can provide better characterization than laboratory distillation or simulated distillation tests.
     
   • Check with lab supervisors, safety supervisors, production supervisors, health and environmental officers what measures and equipment (e.g., protective clothing, access, training) are needed for sampling and for other tests duties. Order any unavailable items.
     
   • Contact the research department for input. Decide whether the designer should be contacted.
   
   
2. Set a tentative date.
   

   Check with the various departments how long it will take to carry out the preparations for the test. Add a “Murphy’s law” factor and set the earliest date at which the test can be carried out. A good “Murphy’s law” factor is half the longest preparation period, or 1 week, whichever is longer. For instance, if the instrument department advises that it will take four weeks to calibrate all instruments, allow another two weeks for delays. Better Murphy’s law factors can be determined by experience in each situation.

   
   

   Once the preparations are completed, set the time most convenient for the plant operation personnel. Stable operation during the test is essential, and the operation personnel can anticipate periods of potential instability best. An efficiency test usually means increasing the lab workload, and the lab supervisor will need to schedule additional personnel. Obtain a commitment from the lab supervisor to complete the analysis by a reasonable date. The longer a sample sits on the shelf, the lower the reliability of its analysis because of possible leakage or chemical reaction and the higher its chance of being lost.

   
   

   Inform all parties involved about the proposed dates and duration of the test. Check for scheduled maintenance work or unusual feed conditions that can interfere with the test. The author has seen cases when minor maintenance such as filter cleaning interrupted a test. Do not forget to inform upstream and downstream units and check for potential bottlenecks.

   
   

   Troubleshoot for any linked operations that may adversely affect the test. In one case (444), the reboiler heating liquid also heated rail cars. High heating loads on the rail cars would rob the reboiler of some of its heating duty.

   
   
3. Become familiar with the data acquisition system.
   

   Check the reliability of the data acquisition system and to what extent you can use it without interfering with normal operation. Check the time intervals at which key readings are updated and ensure they fit the test needs. Compare acquired data to control board readings. Investigate any discrepancies. When more than one system is used to acquire data (e.g., a distributed control system and a separate on-line process analytical system), the system clocks should be synchronized and any offsets noted so that data acquired from different systems can be matched to each other.

   
   

   It is frequently possible to have a control computer or data acquisition system programmed to prepare a special report of data, to store key data points more frequently, to trend key variables, to add to its records normally unmonitored data, and to calculate deviations and standard deviations for selected time periods. These options should be considered and utilized as needed.

   
   
4. Prepare data collection sheets for the test.
   

   For readings of computer data logs, it is a good idea to offload 1-minute data from the computer and to later transform them into an Excel spreadsheet, which calculates the maximum, minimum, average, and standard deviation for each measurement over the test period or selected increments of the test period. For other readings, it is best to use the modified versions of the log sheets used by the shift personnel. This is particularly important if the shift team is requested to take readings over the night shift. A new data sheet may confuse them and reduce their cooperation.

   
   

   Readings such as levels, control valve positions, and times at which readings are taken may not appear important but later turn out to be valuable in analyzing results. For instance, a reboiler control valve setting can provide useful information for determining whether the steam side is flooded with condensate. These should be included in the data collection sheets.

   
   

   Include ambient measurements (temperature, barometric pressure, precipitation). They may be valuable for test result interpretation. Case 25.1 in Ref. 201 describes how the correct barometric pressure was central to vacuum tower evaluation.

   
   

   Aim to include as many sets of data as possible on each data sheet to reduce the volume of paper. An attractive flow sheet enhances cooperation from those taking readings.

   
   
5. Carry out a preliminary test run.
   

   This validates the items on the data sheets and identifies the measurements that were left out. It is too late to discover those during the test. Eliminate any unnecessary readings, but be careful in doing this. For instance, a pressure gauge reading that appears unnecessary may turn out to be the deciding factor between two conflicting readings. Do temperatures, pressures, and flows move in sensible directions? If not, investigate and resolve.

   
   
6. Based on the preliminary test run, perform a mass balance on the unit and each column.
   

   Ensure that these close within ±5% (4). Leaking valves and exchanger tubes affect the mass balance, and these leaks need to be identified prior to the final test. Meters have read incorrectly even when the instrument technicians were sure they were right. Ensure that density corrections are adequately applied for all flow meters; correction equations are described elsewhere (4). Reference 361 has superb guidelines for and listing of flow corrections for a wide variety of flow meters. Verify the accuracy of flow meters; use the orifice diameter and pressure drop across it to calculate the flow using the orifice equation as described by Summers (465). Ensure the valves on the lines connecting the orifice taps to the flow meters are correctly set. A material and component balance closure within ±3–5% and an energy balance closure within ±5% are advocated for the final test (4, 187, 192).

   
   

   Some experts (465) accept closure within ±10%, while preferring ±5%. Others (153, 447) target a stricter criterion of ±2% for the mass balance, but recognize that this may not be achievable. When the closure of any of the balances exceeds ±5%, the author strongly recommends performing a sensitivity analysis to define the impact of the nonclosure on the efficiency and only accepting the data when the nonclosure makes little difference to the result or when there is an independent verification of the key meter readings. Sensitivity analysis is discussed in steps 10–14 in Section 3.1.6.

   
   

   In one case, the energy balance on a giant tower closed within ±8%. The sensitivity analysis found this error to be too high to provide a suitable basis for the intended revamp. The plant process engineer solved the problem by diverting the reboiler condensate into a large rectangular trash bin (“Dempsey dumpster”) and timing its filling to give the condensate flow rate. After a couple of trials, the engineer got repeatable results, which accurately matched the steam orifice meter. This test successfully established the reboiler duty as the correct basis for the simulation and revamp.

   
   

   Pay attention to flow measurements of streams near their bubble point or dew point. Flashing of liquid or condensation of vapor can lead to major errors. Any “noisy” or heavily dampened flow meter signal may indicate unexpected flashing or condensation.

   
   

   Whenever practical, cross-check the mass balance against direct weight or volume measurements (e.g., from product tank levels). Make sure you can determine the density of liquid in the tank; you would most likely need to have temperature and composition measurements of the tank liquid. Table 3.2 is a checklist for troubleshooting the nonclosure of material, component, and energy balances.

   
   

   Follow the guidelines in the first two points under step 1 above. Check relief valve bypasses and drains for leakage. Use a pyrometer to find whether these are warm. Leakage will affect the material balance. If it cannot be prevented, check whether it can be estimated from flare or vent header analysis for certain components. Adjust the nitrogen flow to the flare or vent header accordingly.

   
   
   
   
   

   Are any valves leaking (check vents, drains, bypasses, crossovers, relief valves)? Are there any signs of exchanger leaks (cooling water or steam chemicals in the process, process components in the cooling water or steam condensate)? Are any exchanger or control valve bypasses open or leaking? Would any of these leaks negatively impact the mass or energy balance? Are all lines connected to the tower accounted for (look for purges, pressure-equalizing lines, liquid to/from pump seals, relief valve bypasses)? Are the meters included sufficient to entirely define unit performance? Are any handheld meters needed? Are the orifice plates of the correct size? Are their flow factors correct? Are the orifice plates in good condition? Are any orifice taps plugged? Are the valves on the lines connecting the orifice taps to the flow meters correctly set? Are there any flow meters measuring two-phase flow? Can any condensation occur in measured vapor streams? Can any vaporization occur in measured liquid streams? Is there sufficient pipe run free of obstruction upstream and downstream of all flow meters? Are there any spiral wound (seamed) pipes used? Are they discontinued in the vicinity of all meters? Are the appropriate density corrections applied to adjust the flow meter reading? Are all meters zeroed and calibrated? For mass measurements from tank level gauging, are tank liquid temperature and composition metered to determine liquid density? Do flow measurements match volume or mass measurements? Are metered flow rates consistent with those derived from control valve settings and pump curves? Is an ultrasonic flow meter needed? Is there a suitable and accessible measurement point? Is the reboiler return or vapor feed blowing into any pressure tap? Are all the needed samples included? Can all samples be withdrawn in compliance with HSE requirements? Are all the sample lines operable? Any blockages?a Are the sampling procedures adequate?a Can there be any liquid in vapor samples or vapor in liquid samples?a Are the correct sample containers used? Are they adequately purged?a Do the sample containers leak?a Is the preferred analysis method(s) used?a Are the lab analyses correct?a Do any important components fall outside the lab analyzer measurable range?a Are the on-line analyzers calibrated and read the same as the lab?a Did any on-line analyzer produce any nonsensical values in the last three months? Why?a Are the analyzers functioning properly? Are the sample handling loops adequately purged?a Does a log–log plot of distillate/bottom components’ concentration versus relative volatility produce a straight line or a smooth curve?a Does a reaction or accumulation occur in the column?a Can any samples be taken from the column?a For a partial condenser, does a flash calculation match the measurement?a Are the subcooled condensate and superheated steam temperatures measured?b Is an exchanger duty calculated using a small temperature difference (< 10°F) between inlet and outlet streams?b Can the cooling water flow rate be measured with an ultrasonic meter?b Are the air temperatures into and out of the bays of an air condenser and fan pressure drop measured?b Are temperature gauges accurate?b Do thermocouples adequately fit in the thermowells? Are there deposits in the thermowells?b Is feed enthalpy based on the temperature measurement of a two-phase flow?b Does the steam trap leak?b Is there a possibility of water in the saturated steam inlet to the reboiler?b Do heat losses from the column need to be allowed for?b Are the ambient conditions recorded? Are all the HSE requirements (e.g., protective clothing, access, training) needed for sampling and for other test duties taken care of?

   
   

   ^aCheck for separation quality and nonclosure of component balances only.

   
   

   ^bCheck for nonclosure of energy balances only.

   
   

   All unmarked items affect mass, component, and energy balances.

   
   

   A flow meter on a liquid stream leaving the column or drum can sometimes be checked by heavily throttling the control valve on the stream for a measured short time period and timing the level rise in the column or drum. The mass accumulated in the tower or drum can be calculated from the level rise and can be used to check the change on the flow meter.

   
   

   In case of a gross mass balance nonclosure, compare the measured flow rates to those calculated from control valve opening, pump curves, and/or ultrasonic flow meters. If practical, perform this comparison over a range of flow rates rather than a single flow rate. These checks are not accurate, but can often identify the meter that is grossly out of line.

   
   
7. Based on the preliminary test run, perform component balances on the column.
   

   Component balances for the main components, as well as other components for which the separation in the column is important, should close within ±3–5% as prescribed in step 6 above.

   
   

   For minor components, the lower the concentration of the component, the higher the percentage error in the component balance is likely to be (4), which will easily exceed the 3–5% closure. At low concentrations, these components are difficult to analyze, and they may azeotrope with others. Usually, these small components do not affect efficiency determination, and one can get away with even large nonclosure of their component balances. The author encountered situations where even a factor of 2 or 3 in the component balance of a minor component made little difference to the column and its efficiency determination.

   
   

   In cases where the split of the minor components is important, the AIChE Column Testing Guide (4) suggests estimating their concentration from downstream units, where they are concentrated and can be analyzed more reliably.

   
   

   Other cases where the component balance closure may be relaxed are pseudo-components of petroleum fractions, or dilute components, where a 10% closure is considered realistic (4).

   
   

   A useful check for component balances (4, 192) is preparing a log–log plot of the ratio of distillate to bottoms concentrations for each component against its relative volatility in relation to the heavy key component. Relative volatility is calculated under average column conditions. Except for columns that are primarily stripping or rectifying, this plot should give a straight line or a smooth curve approximating a smooth line. This plot can help resolve discrepancies in component balances.

   
   
8. Carry out an energy balance.
   

   An incorrect reflux meter or a leaking steam trap will not show up in the material balance. Ensure closure is within ±5% (4, 166, 192), but see step 6 above for a possible exception.

   
   

   Ensure that the temperatures of subcooled condensate leaving the reboiler and of superheated vapor (or steam) entering the reboiler are monitored. It has been stated (278) that this is critical for reliably establishing column internal flows. Often, there is no temperature indicator on either. Either add indicators or measure using a pyrometer.

   
   

   As a general rule, plant steam is inherently wet (399, 400). The common belief that letting down steam from a higher to a lower pressure through valves can create saturated (or even superheated) steam is often fallacious. Risko (399) shows that letting down 5% wet steam from 200 psig to 50 psig will only reduce the wetness to 2–3%. Audit the steam line to the meter for sources of condensate such as poor condensate trapping (check all steam traps), damaged insulation (check with a temperature gun), and other heat sinks such as uninsulated flanges. Be on the watch when saturated steam or vapor is used to heat a reboiler or preheater. Any water present in the steam will generate a significant error in the heat duty. Check whether the steam temperature can be raised for the test to prevent condensation or whether the condensate can be trapped ahead of the flow meter.

   
   

   For reboilers heated by heat transfer fluids, check the system for contamination, leaks, and degradation products. These may affect liquid physical properties (115), introducing errors in heat duty calculations. Seek guidance from the fluid supplier as needed.

   
   

   Pay special attention to exchanger duties calculated from small temperature differences (e.g., from the cooling water flow and the difference between inlet and outlet temperatures, where the temperatures are less than 10°F apart). Often, the flow can be throttled to increase the temperature difference; if this is impractical, high-accuracy temperature indicators and/or pyrometer checks may be required. Often, cooling water flow measurement is unavailable, but can be achieved using an ultrasonic flow meter.

   
   

   For air condensers, measure the air temperatures into and out of each bay, as well as the pressure difference across the fan. These measurements may prove crucial for evaluating and checking the condenser duty and tower heat balance.

   
   

   For flashing feeds, determining feed enthalpy from feed temperature is often inaccurate, particularly if their boiling range is narrow. Further, the pressure at the feed temperature measurement point is often unknown due to the pressure drop between that point and the column feed zone. A better estimate can often be obtained from the temperature at the last point where the feed exists as a single phase (prior to flashing). This procedure affords an additional validation check when the feed prior to flashing is at its bubble point (e.g., when it comes from the bottom of a previous tower). Using its temperature and composition (per component balance in step 7 above), the bubble point pressure can be calculated and compared with the measured bottom pressure in the previous column.

   
   

   If the feed is preheated, calculate the preheater duty from the heating side (steam side duty or, if liquid heating, the temperature difference between the inlet and outlet temperatures multiplied by the specific heat and multiplied by the flow rate). Similar considerations apply to feed precoolers.

   
   

   If the temperature of a vapor–liquid mixture needs to be measured, an effort should be made to measure pressure as close to that point as possible. Alternatively, the temperature measurement should be as close to the column as possible where the pressure is known.

   
   

   Use a thermal camera to identify any regions of major heat losses. In some high-temperature vacuum columns, columns with small (<2 ft) diameters where the surface-to-volume ratio is high, or uninsulated columns, or where insulation is damaged, radiation and convection heat losses may be significant. In these cases, heat losses need to be estimated or measured and accounted for. In most cases, these losses are negligible (4).

   
   

   Check temperature increases along lines that are heat-traced. This can be achieved by cutting holes in the insulation at the beginning and end of the line and measuring the temperature difference using a pyrometer. There have been instances (240) where heat tracing caused flashing in a reflux line.

   
   

   Check for steam leaks through reboiler and preheater steam traps. These can be detected by comparing the temperatures entering and leaving the trap with a pyrometer.

   
   

   Heat-integrated towers should be evaluated simultaneously for proper checking of the energy balances (465).

   
   
9. Prepare the simulation program and spreadsheet for carrying out the mass, component, and energy balances.
   

   Once they are ready, data can be quickly processed and inconsistencies easily detected before or during the test, before it is too late.

   
   
10. Check whether the tower operates near a pinch or an azeotrope. See steps 10–14 in Section 3.1.6. The efficiency of towers operating near a pinch, or right near an azeotrope, cannot be reliably determined. Efficiencies determined under pinched conditions can be out by a factor of 2 of 3. If close to a pinch, consider increasing the reflux ratio. If too close to an azeotrope, consider reducing the reflux ratio to move away from the azeotrope. For example, if an ethanol–water column produces a 96.1% product, right near the azeotrope, consider reducing the reflux ratio to make a 95% ethanol product. Apply the simulation to guide your move.
    

    If operating near a pinch, a “pinch test” is worth considering. Immediately after completing the main test, the column can be retested at another two or three reflux ratios. The reflux ratios should be higher if no azeotrope and lower if one product is near the azeotrope. If the column is near a capacity limit, before raising the reflux for this test, it may be necessary to reduce the feed rate.

    
    
11. Monitor and enhance the reliability of laboratory analyses.
    

    The AIChE Guide (4) recommends that at least three successive samples of each stream should show essentially constant composition. Below are some actions that help make this happen:

    
    

    A useful trick to monitor lab reliability is to take several samples of the same stream. If possible, take them yourself. Give one sample to the lab tester and request an analysis. Plug the other sample containers, ensuring no leakage. Give the same lab tester the second container several hours later, requesting another analysis (do not forget to remove the plugs first – otherwise the tester may become suspicious). Give the other containers to the lab testers on afternoon and night shifts, leaving one for the tester the next day. Compare the analyses. You may surprise yourself with the difference in results! If necessary, discuss with the lab supervisor, but avoid getting personal. Don’t forget, your objective is reliable data, and bad personal feelings will not help.

    
    

    Another useful check is to obtain a number of test containers early, fill them with samples, then close and plug them to prevent leakage, and record their weight. Reweigh them a number of days later; a weight loss will indicate leakage despite all the precautions. Alert the laboratory supervisor as needed.

    
    

    Check the sampling technique used by the lab tester for each sample. Volatile samples need to be taken in bombs, and these must be adequately purged.

    
    

    Compare the lab analyses against design data, previous performance test results, or by component balances. Look for anything that does not make sense.

    
    
12. Check product analyses using bubble point or dew point calculations.
    

    The tower bottoms stream is at its bubble point, so a bubble point calculation using the bottom composition and the tower bottom temperature should yield the measured tower bottom pressure. Similarly, the tower overhead product should be saturated liquid or vapor at the condensing temperature. Physical property measurement (e.g., liquid viscosity) may also be available and should be checked against the calculated values.

    
    
13. Consider wall temperature measurements.
    

    Holes can be cut in the insulation in strategic locations around the tower, and wall temperatures measured by a pyrometer or a thermal camera. Wall temperatures add invaluable data for validating simulations, testing theories, detecting packed tower maldistribution, spotting accumulation (“bulges”) of intermediate components, and identifying the presence or absence of a second liquid phase. In tray towers, wall temperatures can provide a detailed tray-by-tray measurement of the temperature profile. For rigorous efficiency tests, wall temperature measurements can become the deciding factor (234) between two conflicting models both matching most of the test data. Chapter 7 has many examples in which these wall temperatures were invaluable.

    
    

    Some engineers use contact thermocouples when measuring wall temperature. The author’s experience with contact thermocouples has not been good. They often take a long time to reach equilibrium and are sensitive to surface irregularities.

    
    

    When inserting handheld thermocouples into thermowells, make sure to allow sufficient time to reach equilibrium at the final temperature. This may be lengthy, especially when measuring hot temperatures.

    
    
14. When there is a partial condenser, perform a flash calculation on the reflux drum.
    

    If the calculation and measurements do not match, investigate the reason. The condenser may provide more or less than a single stage. In Case 1.13 in Ref. 201, the capacity of an entire refinery was limited by the gas rate exiting what should have been a single-stage total condenser, but became partial due to Rayleigh condensation providing more than one stage. Should the flash calculation mismatch the measurements, include additional composition, pressure and temperature measurements in the region to fully characterize the split. The author recalls a very painful exercise one time when he skipped this check.

    
    
15. Check feed and draw points where there are multiple valved locations.
    

    The author had been in many situations where the operating team thought the feed entered one point but temperature measurement confirmed it entered another, or the closed valve was leaking and it entered both. A pyrometer survey is invaluable. Then change to the desired feed point.

    
    
16. Check control valve and exchanger bypasses for leakage.
    

    Such leakage will affect the mass and energy balances. Here too a pyrometer survey is valuable.

    
    
17. In deep vacuum towers, attempt to estimate the air leakage rate.
    

    Air leakage reduces the partial pressure of the components, therefore affecting the mixture’s boiling temperature. Attempt to also estimate the point of air entry, as the boiling temperature will only be affected by the leakage from that point up.

    
    
18. Explain the importance of the efficiency test to the shift and lab personnel by informal discussions.
    

    Their cooperation is crucial for obtaining reliable results. Pay special attention to shift supervisors, chief operators, panel operators, and lab testers that will be working with you; they are the key to success. Explain to them the importance of the test, seek their advice, and incorporate their input into your plans.

    
    
19. Tag and label instruments and sample points with weatherproof tags.
    

    Some of the participants are only partially familiar with the test zone. Assign areas of readings for each participant. Request each participant to perform a “dry run” and to point out any instruments they cannot locate.

    
    
20. Prepare a “flag sheet” of the unit.
    

    This is a simplified process flow diagram showing items of equipment, temperature, pressure, and key component compositions on it (Figure 3.3). Use it to take key readings during the test and to troubleshoot for gross deviations. On the sheet, show expected values and next to them leave empty boxes to enter measured values. Perform a “dry run” using this sheet before the test.

    
    
21. Unmetered reboiler steam flow can often be measured by running the condensate to a drum of known volume and timing it.
    

    Check whether this technique can be applied. Remember that steam condensate flashes, giving rise to incorrect readings. It may be necessary to cool the condensate or run it into a drum partially filled with cold water. Check for special safety, health, and environmental precautions associated with such measurements.

    
    
22. Compare exchanger duties with exchanger design.
    

    An exchanger duty grossly exceeding its heat transfer capability based on reasonable coefficients should raise a flag and is indicative of an issue that needs resolving (277).

    
    
23. Ensure sufficient operating margin is available in filters, coalescers, and storage tank levels.
    

    If they reach a limit, they may interrupt the test.

    
    
24. If the material and component balances are to be carried out using a computer.
    

    Preparing the program and simulation at this stage may prove very beneficial.

    
    
25. If involving outsiders such as the designers, consultants, or service providers.
    

    Make sure that they are planning to attend, are aware of the safety training requirements, and are willing and able to comply.

3.1.4 Last-Minute Preparations

Last-minute preparations are critical for a successful efficiency test. Problems that arise at the last minute most likely persist during the test and may lead to meaningless test results. The guidelines below are based on the author’s experience, supplemented by Refs. 4, 47, 121, 192, 286, 319.

1. Verify that all test supplies ordered have been received and are adequately located. Check that the required types and numbers of sample containers and plugs needed are ready at the lab. Tag them with weatherproof tags. These tags should be different from those used for normal analysis. Verify that all the required personnel protective equipment and safety clothing has been received by the participants.

   

   
   
2. Verify that all sample points are unplugged and operable.
   
3. Check that the shift supervisors and laboratory personnel are aware of the test and any special safety, health, and environmental requirements. They will tend to be more cooperative if they appreciate the importance of the test.
   
4. Ensure the availability of a sufficient number of blank data sheets.
   
5. Difficulties encountered during data collection may lower the quality of the data and the cooperation of the participants. Make use of tricks that make data recording easier. These include (121):
   
   
   
   • Supply waterproof dark ballpoint pens for recording data. Inked readings run, while penciled readings are hard to record when the paper is wet.
     
   • Provide each participant with a clipboard for holding the data sheets and taking readings and plastic/nylon covers to protect them from rain.
     
   • Supply light, sturdy ropes for draping the clipboard or handheld devices over the shoulders while climbing ladders.
     
   • Provide pocket notebooks so there is no need to carry clipboards while climbing up tall towers.
   
   
6. Do a final check to confirm that the required instrument, mechanical, and lab personnel are available. Ensure all participants are aware of their duties and of any special safety requirements.
   
7. Do a final check on relief valve bypasses and drains and steam traps for leakage (see Section 3.1.3, steps 6 and 8 and the first two points in step 1). Leakage may have developed since the preparation phase.
   
8. Perform a last-minute check using the flag sheet. Look for any inconsistencies.
   
9. Check that tank levels have sufficient margins and that pressure drops across filters and coalescers are low, so that they do not interrupt the test. Deoil refrigerated heat exchangers as necessary. Inject methanol into columns susceptible to hydrate formation. Backflush strainers.
   
10. Check that there is a sufficient margin between the plant operating rates and the maximum limit. A popping relief valve is the last thing you want during a test.
    
11. Make a last-minute check with upstream units, preferably with the shift supervisors as well as the plant superintendents, and ensure that they are aware of the test. Also check that other units will not constrain the availability of heating or cooling media.
    
12. Postpone the test if an extreme weather condition is approached. Operation personnel are likely to focus on keeping the plant on-line, instability is likely, outside participants will rush to get out of the bad weather, and meters may be adversely affected. In one unit test run (153), cold ambient temperatures (−15 to −30°F) prevented much of the desired data from being collected, leading to poor closure of mass and energy balances and significant errors.

3.1.5 The Test Day(s)

When preparations are adequately carried out, the test should proceed smoothly. The main watchouts in this period are as follows:

1. Ensure that all the safety procedures are adhered to and that protective clothing is worn as required. Unsafe practices are not only hazardous, but can also generate friction with operating supervisors and may lead to the termination of the test.
   
2. At the beginning of the day, verify that the column has been running stably and steadily for several hours.
   
3. Closely watch the flag sheet (Figure 3.3). Look for any major inconsistencies.
   
4. Stay in close touch with the plant superintendent, shift supervisor, panel operators, and upstream unit control rooms. Quickly question any decisions that may affect the test. A nightmare scenario is someone forgetting that the test is in progress and making a move that destabilizes the tower and adversely affects the data.
   
5. Closely watch the sampling technique. Attend personally when key samples are taken. A common issue is the sampler allowing insufficient purge time through the sampling line and sample bomb, especially on stormy days and when sample lines are long. Obtain key samples in duplicate, and keep one in case the other is lost. Check that the sample labeling is correct.
   
6. Ensure that the lab tester can find all the requested sample points. Tests often call for nonstandard samples from spots with which the lab tester is unfamiliar. If the tester cannot find the spot, the sample may come from a wrong spot or not be taken at all.
   
7. It is best to have test readings taken around the clock. If duties are delegated to the shift team overnight, check that they are fully aware of what you want. A phone call to the plant does wonders to avoid problems and to affirm the importance of the test to night-shift personnel.

3.1.6 Processing the Results

The first step is compiling material, energy, and component balances for the unit based on the test data. These should close per the guidelines in steps 6–8 in Section 3.1.3. Verify laboratory analyses using dew point and bubble point calculations. Some flows and compositions may need readjustment to satisfy the balance equations. Any inconsistencies must be resolved before proceeding with result processing.

Prepare a master flag sheet with the test data. This master flag sheet will serve as the test data input to the simulation.

3.1.7 Determining Hydraulic Loads

To perform hydraulic calculations on the tower, tower internals, and distributors, the hydraulic loads are inferred from the simulation. This determination is not always straightforward. The following watchouts apply.

Table 3.3 provides a simplified internal flows output, directly derived from a real tower simulation, for rating trays for a tower revamp. The tower was simulated as 16 stages, numbered from top (stage 1) to bottom (stage 16). There was a major flashing feed entering between stages 14 and 15. In this case, the task was to rate the internals below the feed (the stripping trays) and check whether they could handle the maximum simulated revamp loads shown.

At first glance, the maximum stripping section vapor flow rate is 19,600 kg/h and the maximum liquid rate is 259,600 kg/h. These were the numbers initially sent to the tray supplier. Based on these, the supplier saw no need to change the stripping trays. The client was not happy, as they believed they observed a limitation at current rates and the expected revamp rates were higher.

It is imperative to appreciate that the listing of the vapor and liquid internal loads in the simulation output is often user-unfriendly, especially near the entry point of flashing, subcooled, or superheated feeds or refluxes. Directly using the numbers in the output can lead to incorrect conclusions, as it did in this case. A simple yet effective technique can easily identify the correct liquid and vapor loads: put together a sketch like Figure 3.12 and list the flows on it.

In this tower, each of the bottom two stages (15 and 16) modeled three trays (33% efficiency). The internal vapor flow rate for the upper stage, 19,600 kg/h (Table 3.3), is the vapor flow to the stage, i.e., into the lowest of the three trays represented by stage 15. The flow out of stage 15, which is a lot closer to the vapor flow rate through the upper of the three trays, is a much higher 35,300 kg/h. Table 3.3 lists this flow as the vapor entering stage 14. This is nonsense; the sketch shows the flow into stage 14 to be 35,300 kg/h plus the feed flash vapor (159,300 kg/h), a total of 194,600 kg/h.

Table 3.3 gives the liquid rate from stage 15. This is the flow leaving the lowest of the three trays. Figure 3.12 shows that the flow into stage 15 is significantly higher and equals the flash liquid (262,500 kg/h) plus the liquid descending from stage 14 (12,800 kg/h), a total of 275,300 kg/h. This higher number is not in the computer simulation! It is obtainable from the simulation. You need to flash the feed to column entry pressure and then add the flash liquid to the liquid flow from stage 14 (this number is in the simulation), and the result gives the liquid rate to stage 15. So stage 15 needs to be rated for 35,300 kg/h vapor and 275,300 kg/h liquid.

All these numbers need to be checked with mass balances to ensure the correct numbers are plotted. In this case, the material balance for stage 15 gives:

The author has experience with another major simulation software in which the vapor flash (equivalent to 159,300 kg/h in Figure 3.12) was combined with the vapor from tray 15 (35,300 kg/h) and the result was presented as vapor from stage 15. This would have required stage 15 to be rated for 194,600 kg/h, which is nonsense. In that case, the feed flash was a lot smaller, about 10% of the vapor load from stage 15, so it was missed. The problem was picked by the tray supplier who could not design trays to satisfy the excessive required vapor load. A sketch like Figure 3.12 showed that the additional 10% vapor from the feed flash should not be included in the stage 15 vapor load, and therefore, the supplier’s tray will work.

Caution is also needed with highly superheated feed. The superheat vaporizes some of the tower liquid upon entry. The vapor loading peaks when all the superheat is consumed to generate dew point vapor. The simulation often misses this peak (“the bubble effect”). A classic example is a typical refinery FCC main fractionator, where a highly superheated feed at about 1000°F enters, leaving the stage above it at about 700°F. The maximum vapor rate can be inside the stage and, if not accounted for, may lead to premature flood as reported in one case (286). The author had success by rerunning the simulation with two stages above the feed and using the intermediate vapor rate between them to rate the bottom section. This rerun is only used for the hydraulic loadings, as one stage is realistic for separation.

Similar considerations apply to liquid side draws or pumparound draws. The simulation output may not include the product draw rate, the pumparound draw rate, or both, in the draw tray liquid loadings. Identification of such issues is the same as outlined above: draw a sketch like Figure 3.12 and check the numbers using a material balance as described above.

For packing distributor and parting box ratings, entry piping rating, and pipe distributors bringing feed or reflux, a flash calculation of the feed to tower pressure at the feed entrance point is a necessity. The distributor and piping need to be rated based on the flash vapor and flash liquid rates. The computer simulation often does not report the results of this flash, but would easily include the results upon request by the user. Skipping this flash calculation is a very common gross error in packing distributor ratings. The author has seen many cases where the feed was assumed to be liquid when the flash vapor volume at column entry pressure was more than 80% of the feed. Needless to say, the distributors designed for liquid had no chance of working with a large vapor fraction.

Numbers coming out of the simulation also need to be closely reviewed with the help of sketches like Figure 3.12 with highly subcooled feeds. In one case (113), high ethylene losses occurred after replacing trays by packings in an olefin demethanizer. Inadequate allowance for the highly subcooled (70°F) feed overloaded distributor capacities. In another case (450), inadequate accounting for subcooling in tower revamp design led to premature flooding. In one more case (41), the distributor maximum liquid load exceeded the reflux pump capacity by far. The supplier design was based on the liquid rate from the top stage when it should have been based on the reflux feed to the top stage.

In some applications, vapor and liquid loads change throughout a trayed section or a packed bed. A common optimizing design strategy in packings is to use the higher-capacity, less efficient packing where the loads are high, and the lower-capacity, more efficient packing where the loading is low. In this case, the vapor and liquid loadings at the demarcation point need to be evaluated, with proper allowance for the efficiency change, making sure the lower-capacity packing can handle the intermediate loads.

3.1.8 Hennigan’s Rules

Below are rules of thumb proposed by Hennigan (169, reprinted courtesy of S. Hennigan, INEOS Acetyls) for critically reviewing tray efficiency determinations. While not all may apply to a specific situation, they are all worth considering. Rules of thumb should always be taken for what they are: rough criteria that reflect the experience of practitioners. They are not meant to replace detailed calculations. Their merit is in providing troubleshooters with practical insight that can be useful in their endeavors.

• Rule #1 – For a new system: expect and design for 40–70% typical efficiency for a system around the middle of the O’Connell curve (Figure 2.39) (i.e., systems of moderate relative volatility around 1.5–3 and liquid viscosity between 0.1 and 2 cP).
  
• Rule #2 – Don’t hope to get much more efficiency by modifying a hydraulically sound tray design across its operating envelope.
  
• Rule #3 – Check efficiency in depth, particularly of older kit – they may reveal a performance deficit that has been lived with for years.
  
• Rule #4 – If tower performance “feels” wrong, this is probably a strong indication of an efficiency issue (which may have arisen from emergent hydraulic problems or bad design).
  
• Rule #5 – Measuring efficiency consistently before and after changes and shutdowns enables sensible conversations between vendor and operator.
  
• Rule #6 – Follow through the actual financial consequences of both low and high efficiency – the answer may not be what you think.
  
• Rule #7 – Look for similar efficiencies in families of similar separations. This is the most powerful way to detect deficits and paves the way for future changes.

3.2.1 Troubleshoot for Process Leaks

The first step when evaluating tower separation, and when preparing for efficiency testing, is to eliminate any process leaks (Section 3.1.3, step 1). Any such leaks are likely to misguide troubleshooting and efficiency determination. More importantly, process leaks can be health, safety and environmental hazards. These hazards are dealt with in safety texts and are beyond the scope of this book. Here, we focus on the effects of process leaks on separation and efficiency testing.

Consider also heat exchanger leaks (see Section 7.3.5) and other extraneous materials that may affect tray or packing efficiency. For instance, seal oil leaking into the feed of a heavy water distillation tower covered the tower packings with a thin hydrophobic film, which reduced wettability and halved efficiency (70). In another case experienced by the author, gas oil leaking through the seals of a diesel product pump got the diesel off-spec.

Closely watch out when one of the heat transfer streams is also the product stream. For instance, a reboiler or condenser tube leak may affect the concentration of water in one of the products, as reported in one methanol–water column (333). In another case (284), gas dehydration was poor due to feed–effluent exchanger that leaked cold, wet glycol into the heated lean dry glycol entering the top of the contactor. In one other case (138), a kerosene product from a refinery crude tower showed color because a mid-pumparound heat exchanger leaked crude in. Leakage of crude into its preheating stream (tower overhead, pumparound liquid, distillate product) is a common occurrence that contaminates distillates or plugs trays in the tower. In another case (138), a water leak overloaded the condenser, causing a 40% loss of overhead product. A similar case is presented as Case 5 below. Another case is described in detail in Section 1.2. Useful techniques for exchanger leak detection were described by Pugh et al. (386).

Case 1. Off-spec ethylene product (216) The final step in ethylene plant distillation, the C₂ splitter (Figure 3.13), separates the ethylene product from the ethane recycle. This is a close separation under pressure (typically 200–280 psig), which produces a high-purity ethylene product (typically a few hundred ppm ethane impurity) and requires a large number of trays, with about 100 trays common.

At a turnaround, one C₂ splitter was retrofitted with high-capacity trays to raise throughput. Upon restart, the ethylene purity spec could not be met. Tray efficiency was less than half the design. Eventually, the plant got the ethylene on-spec, but at the price of losing tons of ethylene in the bottoms and operating at very low throughput. For an ethylene plant, this is a performance disaster, costing tens of thousands of dollars, maybe more, per day. The plant recruited experts, troubleshooters, and gamma scans. There were meetings, discussions, brainstorming, theories, and ideas. The gamma scans surprisingly showed no apparent problems, but they were not believed.

Two weeks later, an operator climbed up the tower. The C₂ splitter had a differential pressure transmitter (Figure 3.13), mounted right at the top of the tower. This is a good location for minimizing issues with liquid accumulation in the impulse lines (192), but is also an inconvenient spot to look at and maintain. The operator noticed that around the transmitter was a short bypass line with a ¾-in. valve. To his surprise, the valve was open, probably to isolate the transmitter and protect it during some commissioning operation. It should have been closed when the transmitter was reconnected, but this did not happen. The open valve passed a large quantity of ethane from the tower bottom to the top product. A typical differential pressure on this tower was about 8–10 psi, which will push a lot of ethane into the ethylene via this ¾-in. line.

When the operator closed the bypass valve, the ethylene purity came good, the trays operated efficiently, and the consultants were sent home.

The author had another similar experience with an analyzer cross-connection on another C₂ splitter left open. This tower also had problems making on-spec ethylene upon restart.

Case 2. Poor butene separation (216) The author was involved in consulting a client who built a new butene purification facility. The product tower, which was the largest in the plant, separated two butene isomers. Again, a difficult separation. This was a licensed process. Two licensors bid: one offered a tower with 160 trays, and the other with 100 trays. Our evaluation was that 100 trays would be tight, but would at least get close to achieving the desired separation. The large cost difference gave the client sufficient incentive to go with the shorter tower, and its licensor was chosen.

Upon startup, the worst fears materialized. The tower was unable to make on-spec products without increasing the reflux to the point of significantly losing capacity. “We told you so, you get what you pay for” was heard from the skeptics that favored the taller tower. The author was called in to see whether there was anything that could be salvaged.

Learning from Case 1, troubleshooting began with looking for any possible connections between the products. There was no bypass around the differential pressure transmitter. However, to reduce the number of storage tanks, the design provided the flexibility of sending either product to each of the three storage tanks (Figure 3.13). So there were cross-connections. The author checked each valve on these connections. At the time, the top product was going into tank #1, and the bottoms product to tank #3. All valves marked as “NF” or “SHUT” were shut. As the author was checking them, he put his hand on the lines on each side of the valves (the product temperatures were low enough so there was no hazard in doing so; if the lines are hotter, putting a hand on the line is a burns hazard, and a pyrometer should be used instead). One of the cross-connections felt slightly warm. The author called the operator. “Let me give you another half a turn on the valve and see if it helps,” said the operator, but he could only close it another quarter of a turn (see Figure 3.14).

The next day, the operators told the author that overnight they comfortably reached the design throughput with the products on-spec. The reflux was still slightly higher than the design, but not enough to bottleneck the tower capacity.

Case 3. Loss of yield For years, engineers in a refinery wondered why the diesel yield from the atmospheric crude tower fell short of the simulated value, while the gas oil product yield was higher. Eventually, a process engineer decided to troubleshoot for cross-connections between these products. She identified some pumps that could be used to pump either product, giving flexibility to perform maintenance on the pumps. So there were cross-connections. She checked the valves on these cross-connections; all were shut. She then took readings using a pyrometer. One of the cross-connection lines was hot. Closing the valve another half a turn did not help here. However, that line could be isolated and the valve replaced. Following the valve replacement, the diesel yield predicted by the simulation was achieved.

Case 4 Sampling from a different point (Case 8.1 in Ref. 201, contributed by Mark Pilling). Kerosene product was off-spec with a high end point. The cause was identified by taking a sample immediately off the column rather than from the product rundown line. This sample was not taken initially as it was hot and awkward to obtain. The sample directly off the column had the kerosene on-spec. Pursuing this lead showed a leaking valve between the diesel and kerosene products upstream of the normal product sample point. This leaking valve was repaired, and the column achieved the design rates and product specifications.

Case 5 Sampling from a different point detects exchanger leak (432) An amine absorber unit (Figure 3.15) was revamped with high-capacity trays. Following startup, the absorber failed to provide the same removal of H₂S that was achieved prior to the turnaround. The initial investigation, focused on the new high-capacity trays, did not uncover any issues.

A lean amine sample from a point immediately downstream of the regenerator showed good quality. Later, an operator took a sample of the same lean amine stream but from downstream of the cross-exchanger (Figure 3.15) and noticed a slightly different color. He had it analyzed. The sample showed contamination with rich amine. The leak raised the H₂S concentration in the lean amine, leading to excessive H₂S in the treated gas. An examination of the exchanger in the next outage showed pin-sized holes in many of the tubes.

Takeaway from all these stories is that the very first step in any troubleshooting of poor or substandard separations, or in planning for a tower efficiency testing, is to critically look for leaks that can impair product purity across all cross-connections and heat exchangers.

3.2.2 Troubleshoot for Tray Weeping

Weeping refers to liquid descending through tray perforations. Some liquid may also leak at the joints where tray sections are bolted or clamped to supports. Liquid descending through the perforations or at the supports short-circuits the primary contact zone, reducing tray efficiency.

At the tray floor, the static liquid head tends to push the liquid down the perforations. The vapor (dry) friction pressure drop counteracts the downward force, acting to keep the liquid on the tray. Weeping occurs when the static liquid head exceeds the dry pressure drop. For most trays, some weeping takes place under all conditions. Sloshing and oscillations create localities in which the static head exceeds the dry pressure drop, inducing intermittent local weeping.

Some weeping, as much as 20% (106), can be tolerated without an appreciable loss in efficiency (74, 106). Tray weeping may be nonuniform (17, 300). Liquid weeping near the tray inlet drops to the outlet of the tray below, bypassing the contact zones of both trays, making it much more detrimental to tray efficiency than uniform weeping. Liquid weeping near the tray outlet drops to the inlet of the tray below, bypassing very little of the contact zones, with less efficiency loss than uniform or inlet weeping. The maximum amount of weeping that can be tolerated therefore varies with the system and with the location of weeping on the tray, with typical values between 10% and 30%. Excessive tray weeping significantly reduces tray efficiency.

Maldistribution of liquid and/or vapor to multipass trays can also induce nonuniform weeping. In addition, in two-pass and multipass moving valve trays, multiplicity of steady states (348) may create another type of nonuniform weeping, in which some panels weep and others do not.

Amazingly, in some polymerizing services where the polymer forms on dry surfaces, a mild amount of weeping can be beneficial. In such services, a polymerization inhibitor is usually injected into the top reflux, protecting the wet upper side of the trays. If the underside of the trays is unwetted, the inhibitor does not get there, and the polymer will accumulate underneath. In this case, it is beneficial to allow some weeping, which will keep the underside wet.

Hydraulic calculations are the best methods to evaluate weeping. For lower-pressure systems (<150 psig), the author and others (74) have had an excellent experience with the method of Lockett and Banik as modified by Colwell and O’Bara (74). The author has also had good experience with it in high-pressure non-distillation systems like amine absorbers (223). For high-pressure distillation systems, the author and others (74) prefer the Hsieh and McNulty correlation (179). Both of these correlations and their ranges are presented and discussed in detail in Perry’s Handbook (245) and Distillation Design (193).

The above correlations are based on uniform weeping and do not account for intensification due to nonuniform weeping. In addition, the above correlations make no allowance for the intense inlet weeping generated under channeling conditions as described in Section 2.12, which is detrimental to tray turndown. In one air separation column with sieve trays at 160 mm tray spacing and excessive open area, the products went off-spec at rates below 95% of the design (514).

In some services, corroded tray support rings are replaced by an internal support structure in the tower. The author had an experience with extensive weeping with this technique, which is unaccounted for by the above correlations.

When performing the calculations, audit the uniformity of vapor load in the section under consideration. In some applications, such as steam stripping of hydrocarbons or organics, there is a large variation in vapor loads. This is illustrated in Table 3.3, where the internal vapor flow rates in the stripping section of a refinery crude tower varied from 4500 to 35,300 kg/h, or from 4158 to 8300 m³/h. On a C-factor basis (Eq. (2.1)), this is a variation by a factor of 4–5. If the trays in that section are all of the same design, weeping will be intense in the lower trays. A successful strategy to minimize weeping is to progressively reduce the tray open area from top to bottom in this stripping section (156, 165).

Another useful concept in weeping evaluation is the Summers et al. stability factor (468) defined by Eq. (2.6). When the stability factor exceeds 0.6, weeping is less than 30% and the tray is stable. Weeping and pass maldistribution rapidly escalate at lower stability factors, with 50% weeping or more by the time the stability factor declines to 0.4. Recently, Summers (467) analyzed Fractionation Research Inc. (FRI) data to extend the range of physical and mechanical parameters in the stability factor equation. His revised elaborate equations are presented elsewhere (467). Equation (2.6) is still a good first approximation, but the revised stability factor can have values lower than 0.6 for high-pressure distillation systems and higher than 0.6 at low operating pressures. The author has extensive good experience with the earlier version, but only limited experience with the updated correlation.

To test for weeping, efficiency testing can be repeated at turndown, and the efficiency difference evaluated. Some towers are constantly operated at low rates with poor separation. If excess weeping is confirmed by hydraulic evaluation, these should be tested by increasing reflux and reboil. Separation should improve, as it did in Case 4.4 in Ref. 201.

Leakage, distinct from weeping, refers to the passage of liquid from openings in the tray that are not intended for vapor flow during normal operation. These include (448) leakage around tray support rings, along joints in the tray deck, and through weep holes. Leakage is usually insignificant when a tower operates at high rates, but it may become significant at turndown. It is also important in downsized columns that have been retrayed for internal vapor loads much lower than the original. Seal welding or gasketing and close inspection (Sections 10.3.6 and 10.3.7) are key. Reference 445 contains useful guidelines.

To distinguish tray weeping from other types of internal leakage, the vapor and liquid sensitivity tests described in Section 2.7 may help. The difference is that here one is not looking for flood but for tray efficiency change, so the tests need to be run long enough to afford a reliable efficiency determination. This makes the period of off-spec product longer than in the flood sensitivity tests described in Section 2.7. Regular tray weeping is sensitive to vapor rate but generally insensitive to liquid rate. So for regular tray weeping, the vapor sensitivity test would show a significant efficiency loss, while the liquid sensitivity test would not.

In refinery fractionators, there are often other indicators, such as internal and side product flows, that can be used to monitor the vapor–liquid sensitivity test, circumventing the need for efficiency determination. A good example is Test 4 in Section 9.2.5.

3.2.3 Diagnosing Side Draw Liquid Starvation

A mass balance around the side draw in Figure 3.16a gives

(3.1)

where D, L, WEEP, and SUMP LEAK are all mass flows in consistent units.

The side draw will be starved either when the liquid flow rate L from tray 11 above is too low and/or when WEEP from the draw tray (tray 12) and/or SUMP LEAK is excessive. A low liquid flow rate from above can be due to a condenser limitation. In refinery fractionators, it can result from heat transfer limitations in an upper pumparound circuit (e.g., due to exchanger fouling) and/or from excessive condensation in a lower pumparound circuit, as described in some case studies (215, 312).

Starving the draw forces reduction in D. The mass balance will compensate by increasing the flow rate of a lower or bottom product stream. This induces lighter material that belongs in the side draw into the bottom or into a lower product, getting them off-spec.

We have experienced high WEEP due to channeling (228, see Section 2.12), and it has also been reported by Martin et al. (312). Starving the draw is a common issue when the draw tray is damaged (e.g., 310) and/or when the draw sump is leaking (e.g., 61, Cases 10.3 and 10.4 in Ref. 201).

The design in Figure 3.16b allows some liquid to flow back from tray 13 to the sump, so the draw will be starved at a larger weep rate. Using leak-resistant trays, seal-welding the draw sump, and replacing the draw trays by chimney trays are measures that effectively circumvent draw starvation (156, 192).

To test for draw starvation, increase the boilup, the reflux (or internal liquid flow) rate, or both. In the absence of severe damage, both should help increase the draw rate. Raising the internal liquid rate test (at a constant internal vapor rate) can help distinguish a draw starvation from an undersized draw line bottleneck. With a starved draw, the product rate should go up. If the line size is limiting, increasing the liquid flow will not raise the product rate.

With a starved draw, gamma scans will show the draw lines are not liquid-full.

Side Draw Connected to Two Tower Nozzles (Figure 3.16c) Some towers mix side draw streams from two different elevations to optimize product purity and flexibility. A possible issue described by Lee (276) is that the higher hydraulic head at the upper nozzle can push backflow through the lower nozzle. Symptoms include (276) poor product composition of the upper draw, same temperature at both draws, lower static pressure at the lower draw than at the upper draw, and gamma scans indicating less liquid traffic between the draws than below the lower draw.

3.2.4 Diagnosing Once-Through Reboiler Liquid Starvation

A once-through reboiler uses a total draw via either a downcomer trapout or a chimney (collector) tray per Figure 3.17a or b, respectively, to capture all of the liquid leaving the bottom tray and feed it directly to the reboiler. The reboiler return is directed to the tower bottom sump, from which liquid is drawn as the bottoms product. No bottoms liquid is recycled back to the reboiler, hence the name once-through.

Once-through reboilers can achieve one full theoretical stage of separation (if the trapout draw does not leak) and can also maximize the reboiler temperature difference. Both are advantages over circulating thermosiphon reboilers, in which the recirculation of hot liquid from the bottoms sump into the reboiler reduces the temperature difference and mass transfer stages. Once-through reboiler application is usually limited to systems in which the boilup is less than about 40% of the bottoms product rate, due to the normal limitation of 30% maximum vaporization (by weight) in thermosiphon reboilers (109, 182).

An inherent issue with the trapout arrangement (Figure 3.17a) is weeping (or leakage) from the bottom tray or draw sump. Such leakage has been identified (198) as one of the two most troublesome malfunctions in reboilers. Numerous case studies were described in the literature and surveyed in Ref. 201, together with some of the author’s experiences. In one case of severe draw pan leakage (61), the bottoms specifications were hard to achieve, lights were lost to the bottom product, and steam consumption was excessive, costing US $1.5 million/year. The leakage problem is most severe during turndown and startup, where the vapor rate is low. There have been towers that could not be started up due to this leakage. When all the liquid bypasses the reboiler, the reboiler will have nothing to boil, and the lack of vapor will perpetuate the leakage.

In most cases, leakage from the trapout arrangement will give a symptom of reboil shortage with lights in the tower bottoms. It is best diagnosed by measuring the bottoms temperature, the reboiler return temperature, and the temperature of the liquid draw from the bottom tray. A flash calculation based on the temperature measurements can estimate the fraction of liquid leaking. If there is no leak, the bottoms temperature will be the same as the reboiler return temperature. The closer the bottom temperature is to the liquid draw temperature, the larger the leak. If one of these temperatures is not monitored, use the wall temperature measurement as described in Chapter 7. The temperatures in Figure 3.17a show a typical case in which the tower bottoms liquid consisted of about 30% of liquid weeping (about 10% of the weeping liquid flashed).

As the trapout provides little degassing time, rundown lines to once-through thermosiphons must be sized for self-venting flow (109, 192). When they are undersized, the trapout pan may overflow and again the overflow will bypass the reboiler.

The chimney tray arrangement (Figure 3.17b) can overcome most, even all, of the above problems but is not totally immune. Leakage can occur through the joints of the tray or draw pan (seal welding or good gasketing can prevent this) or through a damaged panel. When the rundown line is undersized for self-venting flow and the liquid residence time on the tray is less than 0.5–1 minutes, overflow may occur. When the vapor velocity through the chimneys is high, it may prevent the descent of the overflowing liquid and entrain it to the next tray, which in turn may initiate flooding that can propagate right up the tower. Figure 3.17b shows the equivalent temperatures after the bottom tray in Figure 3.17a was replaced by a seal-welded chimney tray. The leakage stopped and the bottom temperature increased, indicating less lights in the bottoms. The reboiler return temperature decreased, showing a larger liquid fraction in the reboiler effluent.

3.2.5 Troubleshoot for Liquid in Vapor Side Draws

Liquid can enter a vapor side draw either by entrainment from the tray below or by weeping from the tray or collector above.

To diagnose the source of liquid, it is useful to test at high and low rates. More liquid in the side draw at higher rates indicates entrainment. More liquid in the side draw at lower rates indicates weeping.

When sampling the vapor draw from pressure columns for liquid, beware of condensation due to the Joule–Thomson effect as the vapor expands from the column pressure to atmospheric. It is best to fully open the valve from the draw line to the sample container and to throttle the sample container outlet valve.

Draw a to-scale sketch of the side draw (Figure 3.18). In Figure 3.18, the bottom of the draw nozzles is 18 in. above the tray floor. Research movies (106; see also Figure 2.1a from this movie) show that when a tray operates at about 50% of jet flood, the spray height reaches 12–15 in. So at higher rates, some of the spray is likely to be entrained into the draws. Case studies of liquid entrainment into vapor side draws were presented (201, 367). In packed towers, vapor draws should be placed beneath a total liquid collector. They must not be placed inside or directly beneath a packed bed as these will draw liquid with the vapor.

Check whether the vapor draw is via a bare nozzle or a perforated header. Vapor approaches a bare nozzle from the sides, from above, and from below, and the flow from below may induce entrainment. In packed towers, this flow pattern may lead to vapor maldistribution. Perforated pipe distributors, usually with upward-pointing holes, are common for vapor draws, especially in large towers, but they need to be correctly designed and have the perforations shielded from liquid weeping from above. Invaluable design and checkout guidelines are available in Ref. 441.

Weeping is another source of liquid in vapor side draws. In one case (450), a revamp of a solvent stripper in which bubble-cap trays were replaced by valve trays led to solvent losses in the vapor side draw, which was believed to be caused by the increased weeping from valve trays compared with bubble caps. In Case 10.7 in Ref. 201, weeping from the weep holes of a collector mounted above the vapor draw led to liquid in the draw.

3.2.6 Troubleshoot for Missing or Damaged Trays or Packing

Damage incidents may occur pre-startup (poor installation), during startup, or during the run. These incidents may be sudden (e.g., due to a pressure surge) or gradual (e.g., due to corrosion). The main indicators of missing trays or packings are historical tracking, pressure drop measurements, and gamma scans. Below are some troubleshooting guidelines:

1. Plot pressure drop measurements against vapor rate, as per Section 2.8, Figures 2.5 and 2.6. Compile a reference plot from the data gathered in the first post-turnaround month, and superimpose later data on the plot. Figure 2.17c indicates damage sometime during the run. A plot like this for discrete time intervals after restart will indicate how sudden or gradual the damage occurred. If sudden, it usually correlates with some upset. This plot also allows close tracking of the events. In some cases, damage may cause pressure drop to rise because the debris block some of the lower trays, downcomers, or packing, initiating flood. The discussion in Section 2.8 around Figure 2.19 and the case described in Ref. 226 are excellent examples of a sequence of steps in which pressure drop first rose due to the blockage of trays by debris from corrosion products and later decreased as the debris were washed away.
   
2. If the damage incident occurred during startup, there would not be a good reference line. It may be possible to compile a reference line from pressure drop data from the previous run, providing the internals were unaltered, the pressure drop was unaffected by corrosion or fouling, and the instrument was not recalibrated during the outage. The author’s experience with compiling a reference line from previous run data has generally not been good. Caution is required.
   
3. A pressure survey, as described in Section 2.6 and illustrated in Figures 2.7 and 2.8, is invaluable. Unlike the plotting in steps 1 and 2 above, it does not provide a time record, but enables narrowing in on the region where the damage occurred. Regions of very low pressure drop indicate the likelihood of missing or badly damaged trays or packings.
   

   Lieberman (287) describes a classic atmospheric crude tower case in which four uninstalled manways resulted in little separation between diesel and residue, inducing diesel into the residue. To boil off the diesel, the refinery added a new 200-MM Btu/h vacuum heater. Lieberman measured a 0.1-psi pressure drop across the four trays where 0.6 psi was expected and advocated repairing the four trays rather than adding an expensive and wasteful heater, but his recommendation was not implemented. In another tower (312), the field-measured 0.1-psi pressure drop across 11 trays diagnosed their damage.

   
   
4. The operation team knows the tower best and will be able to put their fingers on when and how its behavior has changed. They will also be familiar with the tower history and events (some seemingly unrelated) that happened around the time of the damage and could have contributed to it. Incidents from the far past may provide insights which the long-serving operators may recall. All these are invaluable inputs to a damage investigation. Work closely with the operation team.
   
5. Damaged or missing trays or packings perform little separation, often indicated as off-spec products. A small temperature difference over a section of the tower, where normally there is a significant difference, can indicate damage. The loss of separation discussed in Section 2.6, and Figures 2.14, and 2.24, is adaptable to tray or packing damage troubleshooting.
   
6. Gamma scans are invaluable for detecting damage. This is discussed at length with many case studies in Chapters 5 and 6.

3.2.7 Poor Material Balance Control Produces High Impurities

Distillation towers are material-balance-controlled (50, 192, 431, 453), i.e., the material flow in and out of the tower is controlled. For this principle to work, a product stream cannot be on an independent flow control. Note the word “independent.” The flow rate of a stream on flow control, with a cascade (or reset) by another variable such as temperature or level, is not independent; it varies as the temperature or level manipulates it. This will not violate the material balance control. In contrast, a stream on a flow control with no resets has a fixed flow rate and does not allow changing the material balance.

The top product in Figure 3.19 is on an independent flow control. Consider 100 units feed, 50/50 lights/heavies, split into 50 lights top product and 50 heavies bottom product, initial steady state. Later, the feed rate is raised to 120 units, the same composition, i.e., 60 units of lights and 60 units of heavies. The flow controller in Figure 3.19 is set at 50, so it would only allow 50 units to leave in the top product. The other 70 units will attempt to leave from the bottom. But since there are only 60 units of heavies in the feed, the other 10 will be lights, and the bottom product will be off-spec for lights, even with high-efficiency tower internals.

The presence of other controllers may alter the response, but will not solve the problem. If a temperature controller manipulates the steam flow rate and a reflux drum level controller manipulates the reflux flow rate, the temperature drop (due to the lights that are unable to leave at the top) will increase the boilup. The additional vapor will be condensed and returned to the tower by the reflux drum level controller. This will cool the tower, and the temperature controller will again raise the boilup, which will again raise the drum level and so on. In reality, the operator will intervene and increase the top draw. But until this happens, one of the product purities will suffer. A case study by Musumeci et al. (330) provides an excellent illustration of how correctly implementing the material balance control tremendously improved product purity. Other case studies have also been reported (192, 319).

Our malfunctions survey (198) identified violations of the material balance principle to be one of the three leading issues in control system assembly. It should be noted, however, that such violation usually affects the purity of only one of the products, not both. In the above example, the top product will remain on-spec. When the purity of one of the products is not important (e.g., when it is recycled to a reactor without any adverse effects), an independent flow control on a product stream may be acceptable. This is practiced, for example, when a downstream unit is limited in capacity and it is desired to keep a constant feed into it.

Violations of the material balance control principle can be prevented by using the material balance control methods described in Ref. 192 and in distillation control texts (50, 431, 453).

3.2.8 Limitations on Reflux or Reboil Generation

Distillation texts (e.g., 34, 193, 451, 492) teach us that to get a required separation, there is a need to provide sufficient reflux and sufficient boilup. This requirement may seem obvious, but in some multicomponent distillations, this may be difficult to achieve.

A typical scenario is when there is a low-ppm spec (typically <50 ppm) for a volatile component in the bottoms, and at the same time, the feed is either much heavier than the design or deficient in a component of medium volatility. In either case, the bottoms temperature will rise, the reboiler temperature difference will be reduced, and the reboiler may be unable to produce sufficient boilup to strip the volatile component out of the bottoms. Cases 2.2 and 2.3 in Ref. 201 give examples of poor stripping of H₂S from the bottoms of a revamped stripper because the feed was much heavier than the design. In Case 2.3, raising the bottom temperature and increasing the reflux ratio were enough to reinstate good H₂S removal. In Case 2.2, raising the temperature was constrained by a reboiler metallurgical limitation, and it was necessary to supplement the reboil by stripping steam injection.

Excess non-condensables in the feeds, or deficiency of condensable components like light keys, may constrain condensation in the overhead condenser, which in turn may restrict the reflux rate. The insufficient reflux may induce excessive heavies in the overhead product and/or may preclude the generation of enough boilup to remove small concentrations of volatile components from the bottoms product.

Bu-Naiyan and Kamali (51) describe a case in which the bottom product from a hydrocracker debutanizer contained 500 ppmw H₂S (design was <5 ppmw) when operating in the diesel mode with the feed much heavier than the design. A simulation study identified poor fractionation due to insufficient reflux as the root cause. The H₂S concentration was reduced to 10 ppmw by spiking the feed with 2000 BPD of butane (light key, about 5% of the feed) to establish adequate reflux/lift. After the injection, the reflux rate could be increased to more than 6000 BPD, which gave the better purity. A further reduction to <5 ppmw was achieved after another problem (Case 2, Section 8.17) was solved. The author is familiar with other cases in which injection or recycle of light key into the feed solved similar problems.

Excessive preheat or feed cooling can induce similar issues. One debutanizer (286) had 72% C₄’s in the feed. A few degrees extra preheat boosted feed vaporization, a drop in stripping rates, and an increase in the C₄’s content of the bottoms. Cases of a lack of stripping in a stripper due to excessive subcooling at the feed (149, 286, 449) were also reported.

Incorrect instrument readings can mislead and induce operation at insufficient reflux and/or boilup rates. Reference 396, Tale 6 in Ref. 216, and Case 2.6 in Ref. 201 are examples. Energy balances and differential pressure checks can help diagnose. In other cases (449, and in Case 2.1 in Ref. 201), reflux flow was cut out altogether, resulting in poor separation.

Good simulation based on field data is invaluable for detecting all the above issues. In some cases, the simulation shows almost no vapor. In one case (149), temperature difference and pressure drop measurements across a crude tower stripping section showed insufficient stripping steam; doubling the steam rate was a major improvement. In another case (396), pressure drop measurements helped diagnose insufficient reflux and boilup.

3.2.9 Control Instability Increases Impurities

There is a wide variety of control problems other than those discussed in Section 3.2.7 that can also increase product impurities. In most cases, these problems lead to instability and fluctuations in reflux and boilup, which in turn affects product purities. It will take a book to address these problems in a way that will do them justice. So the author is leaving their discussion to Refs. 208, 192, 201, and 206 and distillation control texts (50, 431, 453).

3.2.10 Impurities and Contaminants Affecting Azeotroping and Product Purities

In chemical systems, azeotroping of compounds often determines which direction certain components will travel in the column and in which product they will end up.

The presence of contaminants, caustic, or salts can change the equilibrium, sending components in an undesirable direction. The author was troubleshooting the presence of water in the bottom of a debutanizer. The hydrocarbon feed contained a minute amount of water, which would normally end up in the top, with none in the bottom. The test that diagnosed the problem was letting a sample of the bottoms settle and then inserting Litmus paper into the water phase. The dark blue color indicated caustic. Subsequent analysis showed it to be over 50% weight caustic. The tower feed passed through a dry caustic bed to remove acidic components, and caustic fines carried over into the tower.

Similarly, the author has seen many cases where water was supposed to form a minimum boiling azeotrope with an organic low-boiler and leave from the top. Due to salt or caustic entry, a significant concentration of water was often found in the bottoms product.

3.2.11 Absorption of Sparingly Soluble Gases Affecting Product Purity and Downstream Venting

Upon contact, process liquids absorb small quantities of volatile components or sparingly soluble gases. For instance, when nitrogen is used to keep up the reflux drum pressure, a small quantity of nitrogen will be absorbed by the reflux and liquid product. This can make the product off-spec or generate venting problems downstream.

The contact pattern is key. Liquid splashed into the drum vapor space generates a large contact area between the liquid and the nitrogen, and the liquid will absorb a considerable quantity of nitrogen. In contrast, if the liquid line enters the drum well below the liquid level, the liquid will not contact the nitrogen and the absorption will be minimal (if any).

Figure 3.20 describes a case of nitrogen used to maintain pressure in a newly added debutanizer feed drum (210). Following restart, nitrogen needed to be vented from the downstream debutanizer, deisobutanizer, and depropanizer. The plant noticed that the nitrogen control valve was opening almost fully, suggesting that a lot of nitrogen was coming in. This nitrogen had to get out, and the only open route was with the drum outlet liquid. The high absorption rate was due to the feed entry near the liquid surface. This exposed an extensive contact area between the feed liquid and the nitrogen in the drum vapor space. The author is familiar with several similar cases.

Likewise, a freshwater reflux to a tower will absorb a significant quantity of volatile components, while reflux water generated in the overhead system will be saturated with the light components and absorb much less (471).

3.2.12 Reactions and Contaminants Can Affect Product Purity

In chemical systems, reactions can produce undesirable components that end in one of the products. Contaminants can azeotrope with some components and change the direction of their concentration in the tower. Reactions can also generate components that cause plugging, fouling, corrosion, or foaming. Contaminants soluble in the feed may become insoluble and precipitate out once components to which they have affinity are distilled out of the solution in the tower.

Below are some guidelines that can be useful for troubleshooting reactions that directly (not via plugging or foaming) generate product purity issues:

1. The high temperatures in the bottom of the tower make it a fertile place for reactions. Decomposition and condensation reactions tend to produce light components that end up in the top product. This can be tested by lowering the bottom temperature by about 30–40°F for some time while monitoring the impurity in the top product. Case studies 15.1 and 2.7 in Ref. 201 describe experiences in which temperature-sensitive reactions in the bottom of towers generated light components like water or HCl that contaminated the top product. In one of the cases, the reaction occurred at hot spots in the reboiler. The reaction was stopped by lowering the bottom temperature and eliminating the reboiler hot spots.
   

   Another test is to operate the tower on total reflux for a short period and monitor the impurity. A growing concentration of the impurity in the top product confirms the reaction. In one tower (401), total reflux testing identified a suspected decomposition of highly chlorinated ethanes to be the root cause of poor separation in the bottom packed bed. In case study 15.1 in Ref. 201, operation at total reflux verified that the product contaminant was formed by a reaction inside the tower.

   
   

   When the tower bottom contains thermally unstable compounds that decompose exothermically, excessive temperatures can cause a “runaway” reaction and sometimes lead to an explosion. Some experiences with such explosions have been reported in distillation of nitro compounds, peroxides, hydrocarbon oxides, and acetylenic compounds (10, 86, 87, 255, 309).

   
   

   In services where exothermic polymerization occurs, hot spots may form, with the polymer around them hindering cooling by the bulk liquid. This may intensify the above reactions.

   
   

   In batch distillation, as the light components are depleted from the residue, the bottom heats up, often initiating decomposition and condensation reactions that generate off-spec products and/or corrosive components. A common cure is to first recover the product from the heavies. Often, this is a quick step that utilizes the high volatility gap.

   
   
2. Many reactions are sensitive to acidity. Changing the pH can promote desirable or undesirable reactions. Tight pH control is important. Maintaining sufficient margin from the sensitive zone also helps but is not always practical. One scrubber that used water for absorbing Cl₂ and HCl from off-gas (59) required a pH of at least 3 in the bottom for good absorption. Since pH follows a logarithmic scale, increasing the pH to 3.5 as a margin for upsets required tripling the water rate, which was impractical. The higher pH could only be managed by adding alkali.
   
3. The presence of oxidants or reducing agents promotes undesirable reactions in the tower or in storage. When contacting alcohols, oxidants often form aldehydes which promote foaming and product contamination or corrosive acids. When contacting unsaturated compounds, oxidants can generate gums, tars, polymers, and plugging. In vacuum towers, air leakage can initiate oxidation reactions in the tower with similar consequences. In one case (116, 117), excessive air leakage into a vacuum tower caused polymerization, which in turn led to plugging and tray damage. In another case, it caused corrosion of structured packings.
   
4. Corrosion or polymerization inhibitors may be the oxidizers or reducers causing the undesirable reaction. If an adverse reaction is suspected, the inhibitor can be replaced by a more inert inhibitor as a plant test.
   
5. Catalyst from an upstream reactor may find its way into the column, converting it into a “secondary reactor.” Carryover of solids (e.g., metal fines) from upstream may have an identical effect. Both can produce reactions and contaminants that end up as product impurities. In one case (116, 117), carried-over catalyst, together with excessive air leakage into a vacuum column, polymerized the bottom product and caused plugging and damage to trays. In another incident (255), iron carried over from an upstream column is believed to have catalyzed a polymerization reaction in an ethylene oxide column, in turn causing decomposition and explosion.
   

   If an undesirable reaction is suspected, it is worth checking column products and deposits for the presence of reactor catalyst and metal fines.

   
   
6. The materials of construction of the column or internals may influence reactions in the tower. In one case experienced by the author, the replacement of ceramic random packing by metallic packing induced the formation of a compound that contaminated the bottom product. In water peroxide distillation, certain materials of construction catalyze decomposition reactions (87). In hot potassium carbonate (hot pot) absorption–regeneration service, tests (421) showed that ceramic random packings from different suppliers had widely different reactivities toward the solution. Laboratory tests under simulated column conditions can often identify an interaction between the tower chemicals and the materials of construction in the tower.
   
7. Contaminants leading to product impurities may be generated in storage or during turndown operation due to long residence times. Materials like water, rust, or sludge tend to accumulate in storage and react with the feedstock or in the column, as reported in one case (112). Reactive chemicals left over in the column system from a previous campaign or from commissioning operations may also produce undesirable reactions or contaminants. Reaction products can azeotrope with components in the tower and contaminate products. In Case 15.2 in Ref. 201, reactions at the column feed entry zone generated azeotropes that concentrated in the rectifying section above.
   

   In one amine absorber (11), aldehyde entering the amine system caused the H₂S in the lean gas to be well above the specification. The aldehyde bonded with the H₂S, which inhibited H₂S stripping in the regenerator. An extensive survey of failures in amine absorber–regenerator systems identified solvent contamination by degradation products or glycols to be a major cause of off-spec lean gas (274). Chemical analysis is the key to diagnosing these contaminants, as demonstrated in Refs. 6 and 432.

   
   

   If the contaminant is chemically unstable or explosively reacts with the column chemicals, a detonation may result; some examples were reported (86, 309).

   
   

   Chemical analysis and bench-scale tests are key to diagnosis. If practical, it may pay to derive the tower or plant feedstock from an alternative source for a trial period.

   
   
8. Long residence times, like those associated with large recycles, can induce a buildup of reaction products or reactive components in the system. Recirculating systems like absorber–regenerators, extractive distillation, or solvent recovery systems are particularly prone to such buildups. In one solvent recovery system (88), the solvent slowly hydrolyzed, with reaction products building up to the point of interrupting separation in the solvent recovery column. The cure was periodically distilling the reaction products out of the solvent. The author is familiar with a similar experience where acidic products of solvent hydrolysis extensively corroded a solvent recovery distillation unit.
   

   In glycol dehydration systems, slow oxidation of the glycol into organic acids over an extended period of time caused severe corrosion (78). The reaction products should be either removed or rendered harmless, e.g., by neutralization. In one olefin plant caustic absorber (359), the buildup of reactive components when minimum flow bypasses of a compressor were open was theorized to generate additional free radical chains that intensified fouling.

   
   

   Long residence times in the heated zone at turndown can also promote undesirable reactions like decomposition. Often, the column pressure can be reduced to bring the temperatures down since the condenser and tower are unloaded at turndown.

   
   
9. Reprocessing of off-spec streams or streams originating from other units in the tower can bring in contaminants. The off-spec streams typically contain contaminants such as air, water, and rust. Those from other processes can bring in a wider range of contaminants such as catalyst, oxygenates, acids, alkali, foamers, and halides. The author is familiar with a case in which a halogenated contaminant from a downstream plant hydrolyzed to form acid in an olefin plant water-wash column.
   
10. Excessive concentrations may lead to undesirable reactions at significant rates. An intermediate component may build up in the tower and react by itself or with other components in the tower to form undesirable, corrosive, fouling, or even explosive compounds (199). The buildup can occur in any section of the tower depending on the volatility of the component and its interaction with the main components in the tower. Most buildups occur toward the middle of the tower. When most of the product is liquid, component buildup in the overhead loop is also common. Drawing internal samples from the tower (if valves are available), performing a detailed wall temperature survey (Section 7.1), or running simulations can help detect this buildup.
    

    Increasing reflux to the tower may increase the concentration of a valuable top product component and promote its undergoing secondary reactions accompanied by a loss of product, as has been reported in one case (471).

    
    

    Several explosions caused by inadvertent concentrations of peroxides in a column or reboiler have been reported (87, 309); similar incidents may occur with other unstable compounds (e.g., 86, 184, 255). Hydrocarbon impurities ingressed in the air intake to air separation plants tend to accumulate as solids in the liquid oxygen at the reboiler of the air separation column. When built up sufficiently, they react explosively with the oxygen (86, 308). The hazard is greatest at shutdowns because the more volatile liquid oxygen vaporizes preferentially, concentrating the impurities.

    
    
11. An undesirable reaction may occur only at turndown due to excessive reagent or long residence times. In one case (192), an additive was injected into the feed to eliminate small quantities of aldehydes. At turndown, it also attacked ketones. The problem was overcome by running the column at high rates and shutting it down when product stocks built up. A possible alternative solution would have been to reduce the additive injection at low rates.
    
12. An unexpected compound may show up as one of the normal components in the routine laboratory analysis of a feed or product stream. Similarly, a component may be present in an unexpected molecular form. In either case, the volatility and solubility characteristics of the unexpected component may cause it to distill toward the wrong end of the column. The routine lab analysis will indicate poor column separation. In one case (401), hydrogen fluoride (HF) was present mainly as carbonyl fluoride (COF₂) in the feed to an aqueous wash column. While HF can readily be absorbed, COF₂ cannot, giving an apparent symptom of poor absorption of HF. Routine lab analyses should be periodically checked against comprehensive chemical assays. Attention should be paid to concisely defining the components in the streams entering and leaving the column.
    
13. The presence of even small quantities of water often generates a reaction or affects the course of a reaction and, more importantly, azeotroping of the components.
    
14. In absorber–regenerator systems, the key to good absorption is to properly strip the lean solvent in the regenerator. Insufficient heat, inefficient stripping by the trays or packings, and contamination of the amine with a strong base have been the main reasons for the poor stripping (272–274).
    
15. Some reactions are influenced by column pressure. In one case (86), a compound that was nonreactive when heated at atmospheric pressure detonated when heated at column pressure.
    
16. Material precipitated out of solution or emulsion may initiate an undesirable reaction. In one solvent recovery column (86), explosive-grade nitrocellulose precipitated out of solution and detonated.
    
17. Unexpected reactions often occur in first-of-its-kind processes or those using a newly developed catalyst. Large-scale commercial processes are often derived from batch or semicontinuous pilot plants, where residence times, impurities, and recycles are different. When an unexpected reaction is suspected, it is often useful to compare key variables in the pilot tests with those in the commercial unit.