We publish situations: the question asked, what was put in place, what it changed. It is the surest way to recognise your own.
Our work is covered by confidentiality agreements. The names are given in a meeting, with the consent of the people concerned.
Eight situations, as they arose
Each in three beats, including where the answer was that a quantity stays outside an in-line measurement.
A drying run to a fixed duration
Pharma, solid dosage · the end of drying becomes an observed state
Read the situation
The challenge. Solid dosage site. The drying time is a set point matched to the slowest batch ever seen.
What was put in place. A spectroscopic measurement in contact with the product in the dryer, several quantities followed in parallel, control charts in the room.
What it changed. The end of drying became an observed state rather than an hour in the schedule. And two of the target quantities proved unquantifiable in-line, written plainly in the report. Detecting an endpoint.
An identity check at goods-in
Chemicals and polymers · the check takes place where the material arrives
Read the situation
The challenge. Every delivery waits for the laboratory. Quarantine, meant as a buffer, serves as a permanent store.
What was put in place. An identity check at goods-in, a spectral library extended to the families actually received, a written rule for a non-recognition.
What it changed. The check happens where the material arrives. The real work was keeping the library current over time rather than the instrument. Raw material identity.
Raw material sorting at goods-in
Cosmetics · sorting at goods-in on 100 % of batches and drums
Read the situation
The challenge. A cosmetic emulsion manufacturer. A vegetable butter accepted on its certificate of analysis, conforming batch after batch. Yet depending on the batch, the finished product’s emulsion came out smooth or lumpy, and this was only discovered during manufacture.
What was put in place. NIR identification at goods-in, on every batch and every drum.
What it changed. A sorting criterion more selective than the certificate, applied to 100 % of batches and drums received. The system now runs in routine at the client’s site. Raw material identification.
A crystallisation followed during the operation
Fine chemicals · the trajectory becomes visible during the operation
Read the situation
The challenge. The size distribution is discovered in the laboratory, after filtration. By the time the result arrives the population is fixed.
What was put in place. A measurement in the vessel during the operation, following the liquid phase and the evolution of the particle population.
What it changed. The trajectory becomes visible: a missed seeding separates from a cooling that ran too fast. The bound is explicit. The measurement describes the volume probed. Particle size.
A coating driven by weight gain
Pharma, solid dosage · unit-to-unit dispersion becomes an indicator
Read the situation
The challenge. What is weighed is what was sprayed. The layer deposited stays unobserved, and how it spreads between units with it.
What was put in place. A contact-free thickness measurement in the pan, returning above all its dispersion between units.
What it changed. Unit-to-unit dispersion became an operating indicator. The subject moved from the sensor to the mount: the tolerance on the probe-to-product distance is the real condition of feasibility. Coating thickness.
A blend followed through to homogeneity
Pharma, solid dosage · the blending curve becomes visible
Read the situation
The challenge. Samples at the end of the cycle, result the next day. The set number of revolutions was fixed empirically and never revisited.
What was put in place. A measurement at the blender lid, with statistical criteria on moving blocks: a convergence is followed rather than an absolute content.
What it changed. The homogenisation curve became visible. On some batches it showed that extending the cycle began to undo what had just been done. That result was unexpected. Content uniformity.
A continuous process in an opaque volume
Chemicals · dead zones and regime changes become visible
Read the situation
The challenge. A chemicals plant. The phase distribution in the pipe is out of sight: the reasoning runs on inlets, outlets and an assumption.
What was put in place. A tomography by wall electrodes, which returns a distribution with no multivariate model to calibrate or to maintain.
What it changed. Dead zones and changes of regime became visible. The counterpart is clear: the image gives a distribution, and a content follows only through a calibration, light but real. Electrical tomography.
Contents on fillets of varying thickness
Food · fat and moisture across the whole batch, sorting ahead of smoking
Read the situation
The challenge. Salmon fillets on a conveyor belt, ahead of smoking. Moisture and fat are known from the laboratory, after the batch has passed, and the product height on the belt varies.
What was put in place. A non-contact NIR measurement above the belt, thirty centimetres from the product, over the whole fillet as it passes, with a laser sensor reading the distance at each acquisition. The robustness study covered height variations of up to three centimetres.
What it changed. Fat and moisture stay precise across the whole batch, with no drift, despite the height variations. Fillets are sorted on the line by quality, and smoking runs on more homogeneous batches. NIR spectroscopy.
Where this work sits
| Sector | Type of process | Measurement question | Stage of the method reached |
|---|---|---|---|
| Pharma, solid dosage | Vacuum drying | When to stop, and on which criterion | Deployed in routine, one quantity set aside |
| Pharma, solid dosage | Blender | Homogeneity, and for how many revolutions | Proof of concept concluded |
| Pharma, solid dosage | Pan coating | Thickness and dispersion between units | Feasibility conditional on the mount |
| Fine chemicals | Crystallisation in a vessel | Supersaturation, evolution of the population | Proof of concept |
| Biotechnology | Culture and purification | Composition of the medium during the process | Scoping and choice of technology |
| Chemicals and polymers | Goods-in on incoming materials | Is this what it is announced to be | Integration |
| Cosmetics | Goods-in of a plant butter | A sort more selective than the certificate of analysis | Deployed in routine |
| Food and agriculture | Product in movement | Contents on a flow whose height varies | Proof of concept and robustness study |
| Anaerobic digestion | Opaque pipe | Phase distribution | Scoping |
The last column says how far each project went.
A proof of concept has two valid outcomes
Some of these projects stopped at the proof of concept, and that is a result. The reasons for stopping are always the same.
- The target quantity sits at trace level, and its signal stays independent of the process axis.
- The variability available is too small to build a defensible model.
- The presentation of the sample in-line dominates the dispersion of the spectra.
- The reference method is less specific than it was taken to be.
We have revised our own conclusions downwards from one phase to the next, moving to a batch-by-batch validation. That is what makes the rest credible.
A documented conclusion either way says why, and what it would take for the answer to change. It costs far less than a deployment that will not hold in routine. The conditions a non-destructive measurement holds on.
Five questions to ask about a reference, ours included
- On which process exactly, rather than in which sector. A drying and a compression carry different measurement questions.
- On which matrix. A powder is not a powder. Particle size, bulk density and colour change the signal, and so the feasibility.
- Under which calibration burden: pure-component model, model calibrated on samples, or no multivariate model. That is what will be maintained for ten years. Which calibration burden.
- Is the system still running today, and who maintains it.
And a fifth: what did you set aside on that project.
The names, in a meeting
We then describe the case comparable to yours, process, matrix, technology, what held and what did not, and where the person agrees, we put you in touch. The entry point stays a scoping phase, then a proof of concept whose criteria are written before the trials.
Part of this work has drawn on skills we do not carry alone: sampling theory, design of experiments, pharmaceutical development, validation. See who works on our projects.
A second kind of reference, and this one is checkable without going through us: the regulatory texts every statement on this site rests on, with their exact reference and date.
Describe your process. You will hear whether we have met the same question, and how it ended.
Forty-five minutes is enough to place your case among the ones we know, and to hear what worked in comparable situations. And what did not.