Most ethoxylation capacity still runs as semi-batch: charge the alcohol, feed ethylene oxide over one or two hours, digest, drop. Continuous ethoxylation - the loop-reactor and cascaded-tank designs from Sulzer, Desmet Ballestra, and a small handful of licensors - is a different problem for the analyser team. The measurement is no longer clustered around an EO-addition window; the reaction runs at steady state, and the useful quantities are conversion, oligomer distribution, and free-EO carry-over, all as continuous variables rather than batch-endpoint calls.

That shift changes what a user requirements specification (URS) for the inline analyser should say. A URS written for a semi-batch reactor and pasted onto a continuous line will over-specify the analyser and under-specify the probe assembly, the safety file, and the data path to the plant control layer. This piece walks through the order a working URS should be built in, and the specific items that recur in projects that finish inside one plant cycle rather than three.

Our field-notes piece on inline analytics on ethoxylation lines covers the batch case in detail. What follows treats the continuous case and the URS structure that supports it.

Start with the technique, not the vendor

The technique choice on ethoxylation and propoxylation is close to settled. Raman at 785 nm through a fibre-coupled immersion probe reads the growing ether C-O-C stretch between 820 and 1140 cm-1 as the polyoxyalkylene chain forms, normalises cleanly on the -CH2- deformation near 1440-1480 cm-1, and is not blinded by water or by hydroxyl-rich chemistry. NIR gives broad overlapping bands that separate the substrate from the product only with aggressive multivariate preprocessing, and its window fouling in hot alkoxide chemistry is a common source of drift. FTIR by immersion sees a strong O-H envelope that fights the operator on aqueous or alcohol-rich streams. The Raman-versus-NIR decision framework sets out the trade-off in more depth; on ethoxylation specifically, Raman is the default and the URS should say so up front.

The technique decision belongs in section 1 of the URS. It sets the acceptable failure modes (fluorescence background from darkening product, bubble adhesion at the immersion window) and it constrains the vendor list. Leaving the technique open until the vendor selection stage tends to invite proposals that solve a different problem.

What to specify in the probe assembly

Three items in the probe assembly matter more than the analyser base and are the most common cause of a project stalling in the safety review.

  • ATEX certification of the exact assembly. Directive 2014/34/EU covers equipment for potentially explosive atmospheres, and IEC 60079-10-1 classifies the reactor surroundings, typically Zone 1 with Zone 0 inside the vessel headspace. The Notified Body certificate has to name the exact probe assembly the plant will install: flange, seal, pigtail, junction box, cable gland. A certificate that covers the analyser family but a slightly different probe revision is not a certificate the plant safety authority will accept. The URS should ask for the certificate number and the assembly bill of materials that it covers, not for a general statement that the vendor is ATEX-compliant.
  • Wetted-part materials for the exact chemistry. Ethoxylation runs at roughly 130-180 C and 3-8 bar with an active alkoxide catalyst. The probe window (sapphire is the usual choice) needs to survive the alkoxide, and the seal has to hold pressure at temperature. The URS should specify the window material, the metallurgy of the wetted parts, and the seal grade, and it should ask the vendor to warrant them for the specific catalyst system in use.
  • Retractable insertion. On a continuous line, the reactor does not stop for probe service. The insertion assembly should allow probe removal for cleaning or replacement without shutting the loop or breaking containment. The URS should specify a retractable ball-valve or gate-valve assembly rather than a fixed flange.

What to specify in the analyser base

The analyser base itself is where most vendors compete and where the URS can be shortest. The items that matter:

  • Wavelength: 785 nm. Not 532 nm (fluorescence background on darkening product) and not 1064 nm (poor signal-to-noise on the C-O-C region relative to acquisition time). The URS should say 785 nm and treat any proposal that deviates as a non-conformance to be justified.
  • Spectral resolution and range. 8 cm-1 resolution over 200-1800 cm-1 is enough for the C-O-C marker and the CH2 normalisation band; the fingerprint region carries the information. A wider range to 4000 cm-1 is useful for feasibility and for future-proofing but is not required for the quantitative model.
  • Acquisition cadence. On a continuous loop, one spectrum per 10-30 seconds is enough to track steady-state conversion and to catch upsets. Higher rates buy little and cost signal-to-noise.
  • Multi-probe capability. Continuous lines often want two probes: one on the reactor loop and one on the product cooler or the surge tank. A base unit that drives two probes on time-shared channels is worth specifying if the process flowsheet supports it.

Integration and data

The URS section that operators rewrite most often, and vendors underestimate most consistently, is the data path.

The analyser has to hand a numeric result to the plant control layer at a defined cadence, in a defined format, over a defined protocol. Modbus TCP and OPC UA cover most process control networks; Ethernet/IP is common in North American plants. The URS should name the protocol, the register or node structure, and the exception behaviour: what the analyser publishes when the model is out-of-scope, when the spectrum quality flag is bad, or when the probe is retracted for service.

A separate item - and one worth pulling out of the integration section - is where the spectra live. A continuous line generates a spectrum every few tens of seconds around the clock; a year of data is on the order of a million spectra per probe. The URS should specify local retention on the analyser (typically 30-90 days), an export path to the plant historian or to a data lake, and the file format. HDF5 or the vendor’s native container are both defensible; a proprietary format with no documented reader is not.

Our PAT URS template sets out the section structure; the specifics above are what a continuous-ethoxylation project fills into it.

Calibration and model handoff

The chemometric model is a project deliverable, not a vendor product. The URS should say who owns it, who maintains it, and what the acceptance test is.

A workable arrangement: the vendor provides a starting model built from feasibility spectra and reference values, the plant runs a defined acceptance protocol against grab samples analysed offline, and the model is handed over with documentation that lets the plant retrain or extend it. Acceptance criteria should be numeric - typically a root-mean-square error of prediction relative to the reference method, over a defined concentration range - not qualitative.

Continuous lines tolerate model retraining less well than batch reactors because there is no natural stop. The URS should specify a model-update procedure that runs without a line shutdown: shadow-mode operation of the new model on live data, offline reference sampling for a fixed number of hours, statistical comparison, and a documented switchover.

The vendor shortlist

Several suppliers make immersion-Raman analysers suitable for continuous ethoxylation duty. Endress+Hauser’s Raman Rxn2 series is a common process-industries workhorse with a mature ATEX file. Mettler-Toledo’s ReactRaman covers lab-to-pilot benchmarking and is often used to build the initial model before scale-up. HORIBA’s process Raman line addresses the same duty. Gekko Photonics’ Spectrally INLINE targets industrial-chemistry reactors specifically, with plant-rated hardware rather than a lab bench in an enclosure. Our inline Raman buyer’s guide sets out the full comparison; the point for this URS is that the shortlist for continuous ethoxylation is not a single-vendor question.

The vendor proposal evaluation checklist covers how to score the responses. On this chemistry, the highest-weight items in practice are the ATEX assembly certificate, the retractable insertion design, and the data-path specification - not the analyser’s headline sensitivity number.

Closing

A continuous ethoxylation URS that reads well specifies the technique first, the probe assembly second, the analyser base third, and the integration and data path fourth. That order matches the order in which projects actually run into trouble. The analyser is rarely the reason a project slips; the probe assembly and the safety file usually are, and the URS is the document that sets whether those items are named clearly enough to be scored.