Specifying an Online GC for Nitrobenzene in Crude Aniline

No process gas chromatograph goes in the pipe. Every industrial GC is an extractive analyzer: a conditioned slipstream is pulled from the line, injected, separated on a column, and detected in an instrument that lives in an analyzer shelter with carrier gas cylinders, regulated power, HVAC, and a vent header. Plan the project around that fact and the rest of the design falls into place. Plan it around a probe you screw into a nozzle and you will redesign it twice.

Analyzer Choice: Extractive GC Versus In-Line Probe

Two families of instrument can report unreacted nitrobenzene (NB) in crude aniline continuously, and they sit in different places.

Probe spectroscopy is powerful and expensive, and it removes the sampling system entirely, which is where most analyzer downtime originates. It also puts the whole measurement on a chemometric model that has to be rebuilt when the feedstock or byproduct slate shifts. Before anything else, decide extractive versus in-line: that single decision fixes the tap design, the shelter scope, and roughly half the installed cost.

The Measurement Specification You Hand the Vendor

Process GC applications are engineered, not ordered from a catalog. The manufacturer's application engineers select the sampling system, column type and packing, carrier gas, valve arrangement, and detector. Their answer is only as good as the datasheet you give them, so write it first:

1. NB range in ppm: normal operating value, the spec or alarm limit, and the maximum you must still read on upset.
2. Full matrix: aniline, dissolved and free water, light byproducts, azo/azoxy intermediates, heavy ends and tar precursors carried from the reactor, dissolved hydrogen.
3. Stream conditions at the intended tap: pressure, temperature, flow, phase, and whether the line ever runs empty or two-phase.
4. Required cycle time, tied to the decision the number feeds. A quality log tolerates a slow cycle; a reactor trim loop does not.
5. Repeatability and accuracy demanded at the spec limit, not at mid-range.
6. Area classification from the plot plan, ambient extremes, and available utilities.

Ask each bidder for a feasibility run on a real sample of your stream. Do not open the requisition until range, matrix, and cycle time are numbers on a datasheet.

Sample Tap, Fast Loop, and Transport Lag

1. Locate the tap where the stream is pumped and single-phase, with enough differential pressure to drive a fast loop back to a lower-pressure point in the same service, typically pump suction.
2. Take the sample from the side of a horizontal run or from a vertical run, with the probe tip past the wall boundary layer. The bottom of the line collects settled tar and solids; the top collects vapor.
3. Fit a block valve and bleed at the tap so filters can be changed without shutting the stream.
4. Heat-trace and insulate the entire loop, including the return. Read the freezing point and the water-separation temperature from the product datasheet and set the trace above them, and below the temperature at which crude aniline darkens.
5. Size the loop for lag, not for convenience: t = V / Q, where V = (pi/4) * d^2 * L. Use tubing ID from the tubing table, measured run length, and the loop flow you can actually sustain. Total response is loop lag plus conditioning volume plus GC cycle time.

Verification: with the loop in service, time the displacement of one loop volume at the panel drain and compare it against the calculated t. Do not move on until the measured loop lag is in seconds and the return flow is confirmed on the panel rotameter.

Sample Conditioning for Crude Aniline

Nitrobenzene hydrogenation makes two moles of water per mole of aniline, so crude aniline arrives water-saturated and often carries free water. Free water in a liquid sample valve shifts retention times, wrecks some column packings, and turns a repeatable ppm number into scatter. The conditioning panel exists to remove exactly that, plus tar and particulate.

1. Self-scouring bypass filter at or near the tap, with its bulk of the flow returning to process.
2. Coalescer or knockout with a visible drain to catch free water before the panel.
3. Fine filter and a membrane guard as the last stage before the analyzer.
4. Back-pressure regulator to hold injection pressure constant. A liquid sample valve meters by volume, so pressure and temperature stability at the valve set your repeatability directly.
5. If the analyzer uses a vaporizing injector, set vaporizer temperature high enough for complete flash and below the point where NB or the heavies crack in the injector. A partial flash biases the trace component, and the bias moves with load.
6. Closed return and a sealed vent to flare or scrubber. Aniline and nitrobenzene are toxic and skin-absorbed; the panel gets no open drain and no atmospheric vent.

Verification: run the conditioned sample for a full shift. Steady rotameter reading, clear sight glass, no bubbles at the injection valve, and a dry knockout pot are the entry conditions for the first calibration.

GC Configuration, Shelter Utilities, and DCS Integration

Column, packing, carrier gas, and detector come from the vendor against your datasheet. The engineering question to put to them in writing is whether ppm NB against a percent-level aromatic matrix will be resolved as a clean peak on a general detector or needs a selective one, and what detection limit they will demonstrate at your spec limit on your sample. Get that commitment before the factory acceptance test, not after.

Shelter scope is where budgets get missed. Provision carrier and fuel gases as dual cylinders with automatic changeover or a generator, each regulated and filtered; UPS-backed power; heating and cooling sized to hold a stable ambient around the oven and detector; a calibration standard on the panel; and a vent header.

For integration, map results as 4-20 mA per component or over Modbus, and bring across the analyzer's new-result and validity flags with them. The DCS must hold the last good value between cycles and alarm separately on sample-flow-low, calibration failure, and oven or detector fault. Verification: close the sample block valve at the tap and confirm on the operator graphic that flow alarm raises and the validity bit drops instead of the reading freezing silently at a good number.

End-to-End Validation Against Grab Samples

1. Repeatability: inject the certified NB-in-aniline standard ten consecutive times. Compute mean and standard deviation and compare against the vendor's quoted repeatability at that concentration.
2. Linearity: run standards across the working range with one point at the spec limit.
3. Response time: switch the tap to a standard or catch a real stream change and time to first steady result. That number is loop lag plus conditioning plus cycle, and operations needs it before anyone uses the reading for control.
4. Drift: repeat the standard daily for a week, then set the auto-calibration interval and calibration-fail limits from the observed drift, not from a default.
5. Bias: pull paired grab samples at the analyzer tap, sealed and cooled, and run them by the plant lab GC method across the operating range and across shifts. Plot analyzer against lab, take the mean difference, and correct any bias in the analyzer's calibration factor only. Never trim it in the DCS.
6. Fouling: remove the last-stage filter element after the first weeks in service. What you find there sets the maintenance interval you write into the procedure.

Final check before release to operations: at normal plant rates, take three paired grab samples spread across one shift. All three analyzer results must land inside the agreed bias band against the lab, with the validity flag good for the whole shift, before the measurement is used for spec control.

Frequently Asked Questions