An empirical, statistically-honest approach to designing extraction processes.
Most extraction failures we see in the field share a single root cause: someone picked a pressure and a temperature that sounded reasonable, ran a few exploratory batches, and locked in the recipe. We don't work that way. Every project AltraProcess takes on follows a disciplined four-phase arc — design, fingerprint, optimize, scale — with statistical modelling at every step.
01. Experimental design
Before any extraction is run, we sit down with the client and define the question. What is the target compound? Is the goal maximum yield, maximum selectivity, minimum impurity carry-over, or minimum cost? These priorities determine which experimental design is appropriate.
For supercritical CO₂ extraction we typically vary four factors: pressure, temperature, CO₂ flow rate, and CO₂-to-feed ratio. A central composite design or Box-Behnken design lets us probe the response surface with as few as 15–25 experiments, depending on the number of factors. For subcritical water extraction we add co-solvent fraction (ethanol or CO₂ at elevated pressure) as a fifth factor.
Because extraction yields and selectivities are non-linear functions of multiple interacting variables. Pressure and temperature interact strongly through their effect on CO₂ density and vapour pressure of the solute. A one-factor-at-a-time study finds only the local optimum along an arbitrary path through the variable space — not the global optimum. RSM finds the latter, and proves it.
02. Analytical fingerprint
Every extract from every design point is analyzed in our chromatography lab. For essential oils and volatile fractions, that means GC-MS with NIST library matching (for identification) and GC-FID against certified standards (for quantitation). For non-volatile bioactives — polyphenols, flavonoids, cannabinoids, alkaloids, tocopherols — we use HPLC with photodiode-array detection.
The result is a multi-dimensional fingerprint for every run: total yield, the concentration of the target compound, and the concentrations of co-extracted impurities. This is the raw data that response-surface modelling needs.
03. Response-surface optimization
With the design-point data in hand, we fit second-order polynomial models for each response of interest. The fitted surfaces let us answer the question the client actually cares about: what combination of pressure, temperature, and CO₂ ratio maximizes my target while keeping yield, selectivity, and cost within acceptable bounds?
The optimum is rarely the run with the highest measured yield. Cost of manufacturing is modelled in parallel: a 5% yield gain that requires doubling extraction time or pressure is almost never worth it on the production line. We deliver the optimum that makes commercial sense, not the optimum that wins lab competitions.
The cheapest extraction process is rarely the one with the highest yield. It's the one that balances yield against time, pressure, and co-solvent cost.
04. Scale-up & confirmation
The lab-scale optimum is then translated to pilot and industrial vessels. Mass transfer behaviour changes with vessel geometry, and the same pressure-temperature combination that worked at half a litre will produce different kinetics at thirty. We compensate using one of three kinetic models depending on the raw material:
- Broken-and-intact-cell model for ground seeds, leaves, and bark — classic Sovová framework. Two-rate behaviour with a fast initial phase from ruptured cells and a slow diffusion-limited tail from intact tissue.
- Shrinking-core model for whole fruits and large particles, where extraction proceeds inward from the surface.
- Desorption-limited model for adsorbed-on-matrix systems — decaffeination of green coffee, removal of waxes from oleoresins.
A confirmation production run at the pilot or industrial vessel closes the loop. Yields and analytical profiles are compared against the model predictions, and any residual deviation is documented in the final report.
Typical project timeline
A complete development project — from raw material in hand to a validated production protocol — typically takes between two and four months. The largest variable is the analytical complexity: a simple essential oil project finishes in eight to ten weeks, while a multi-stage extraction with subcritical water hydrolysis and HPLC quantitation of three or four target compounds can extend to four months.
What you receive
At project close-out the client receives:
- The full set of analytical reports for every design point.
- Fitted response surfaces with regression statistics and validation runs.
- A documented standard operating procedure for the optimized process, ready to hand to a production team.
- A cost-of-manufacturing breakdown using the CRM · CUT · COL · FMC decomposition, with payback-period modelling at the chosen production vessel.
- Kinetic model parameters and scale-up correlations for any future expansion.
- If the project ends with an equipment quote from SuperEx, a complete technical file with P&ID, safety documentation (EN 15614-1, EN 13445), and the operator training plan.
Frequently asked questions
How many experiments does a process study need?
A statistically valid design — typically Box-Behnken or central composite — probes pressure, temperature, flow and co-solvent ratio with as few as 15 to 25 experiments.
What do I receive at the end of the project?
The full set of analytical reports for every design point, fitted response surfaces with regression statistics and validation runs, a documented SOP for the optimized process, a cost-of-manufacturing breakdown with payback-period modelling, and the kinetic model parameters and scale-up correlations for any future expansion.
Can you take the process to production scale?
Yes. Lab-scale optima are translated to pilot and industrial vessels using our in-house kinetic models, then confirmed in a hands-on production run. If the project ends with an equipment quote, you also receive a complete technical file with P&ID and safety documentation (EN 15614-1, EN 13445).