The current configuration recalculated for current supplies and every year from 2028 to 2045 under both scenarios. Design, location, batch, demand and facility settings stay fixed; only the grid, heat and chemical-supply factors change, and every year uses the same draws. Select a year in the chart or a row in the table to show it above.
Table: every year and scenario
Spearman rank correlation (ρ) of each sampled input with the footprint, within the sampled ranges. A screening signal: it is not a share of variance and does not by itself show which check improves the decision.
Enter a measured value for one input of this design and the range of that measurement, then add it. Each measurement is added on top of the ones before (nothing is saved to the data), and every step is recorded: the 90% range, whether the best design in Screen designs changes, and which input to check next.
From the same start, with the same number of steps and the same measurement range, compare filling the input with the highest rank correlation first (guided) with filling the inputs in the order the design lists them (fixed). Each filled input is set to its current central value, so this compares how fast the range narrows, not what real measurements will show.
Placeholder factors applied to the same flows; trade-offs to watch when a carbon reduction is chosen.
Compare manufacturing options
Options use the same random draws only for the same quantity: the same data record (an explicit share key, or the same source and range), the reflux ratio and heat loss of one design's own column (also when it moves to another place or year; two designs, even with the same solvent, draw independently), the laboratory module's fit and its transfer bias for one medicine on one equipment class, the common grid and chemical-supply spreads and each region's grid residual. Differences therefore reflect the designs rather than sampling noise. A pairwise verdict is given only for the same medicine, reporting unit, annual demand, facility sharing and surplus handling; it says whether the pair is a same-year choice or a combined scenario across years, and gives the sampling error of its share of draws. Add options from the Assess tab.
Spearman rank correlation of each sampled input with the difference between the two options. Not a share of variance; a screening signal only.
Screen production configurations
Morphological screen: every combination of process design, location and facility variant is run to meet the same annual demand. Batches, sites and the facility base load follow from the demand, so a larger batch or more slots only helps if the demand uses them. A smaller batch in a facility resized for it is kept apart from a smaller batch in the same facility. Laboratory-only routes and laboratory equipment are excluded; feasible combinations are ranked with uncertainty, with how far each lies above the lowest.
Above the lowest median: the row's median minus the lowest median, and that difference as a share of the lowest median. Lowest in % of draws: the share of draws in which a configuration has the lowest footprint among feasible ones; when several are tied in a draw, the draw is split equally between them. Within the tolerance of the lowest: the share of draws in which its footprint is at most that much above the lowest footprint of the same draw. A configuration that is rarely or never the lowest but within the tolerance in at least half of the draws is close to the lowest, not worse; one that is not close but the lowest in at least 10% of draws is sometimes the lowest: a higher median that still wins often. Every row uses the recovered-solvent setting above: carried to the next year (the default) or discarded at the end of each year, which matters most for designs with few batches. Variants of each design and location: Base; Smaller facility, half batch and Larger facility, double batch, where the facility is resized with the batch (base load per site follows capacity, batch × slots per site, and the load per batch follows the batch size); Same facility, half batch, where the reference design's facility runs half batches with its base load per site and load per batch unchanged. Patient-specific products vary the slots per site instead: Fewer slots, smaller facility; More slots, larger facility; Fewer slots, same facility. Every batch takes one slot of equal length whatever its size, because production, cleaning and changeover hours are not modelled (no time data). A variant that the design's batch or slot limits bring back to the base batch and slots is the base configuration again: it is listed once, as Base, with the variants it stands for. Select a row to open it in the Assess tab.
Check a module against reserved meter readings
The shared solvent-recovery module computes the heat to warm the whole feed and evaporate the recovered solvent at the operating pressure, then the electricity for the heating bath, condenser chiller, vacuum pump and standby. Bath efficiency and standby power are fitted on the development group and used, unchanged, to predict the test group. The Assess tab uses this same fit for laboratory recovery equipment.
For each operation of the chosen design: whether a shared module applies directly, applies once its equipment parameters are determined again, or the design needs records or a module of its own, with the data the design supplies and the modelling work. For laboratory solvent recovery the evidence decides: the module is reused directly only when the design records the same evaporator and metering boundary as the development runs and every sampled run lies inside their conditions. If either is not recorded the equipment match is unconfirmed; if runs fall outside the conditions the prediction lies outside the tested range; other equipment or another meter needs a new fit.
Development records fit the module (their scatter around the fit is the run-to-run noise), calibration records set the transfer bias carried into footprints, and test records are only scored; a record that repeats a development run has no role. A new version starts whenever a record changes (a data file loaded, a case opened, a reset); a case file keeps this history.
Records without run conditions are kept for the check but cannot be predicted: the equipment model needs the pressure and the run time of each run. Where the missing conditions were filled with dummy values, the record is fitted or predicted above and marked “dummy conditions”; those results only show that the chain runs and are not evidence of accuracy. The heat the solvent itself needs (warming every solvent fed from 20 °C to its boiling point and evaporating the recovered share) is shown only as a check against the meter reading. The rest of the reading is equipment overhead (bath standby, vacuum pump, chiller), which depends on run time and equipment power.
Built-in evidence and scenario database
Every input carries a range, a label (measured, simulated or proxy; calculated or assumed for module records) and a status. Demo records are shown as “demo · label” so they cannot be mistaken for real evidence. Labels guide the screening advice: a proxy with a strong rank correlation is a candidate for a supplier enquiry or a measurement.
Load a JSON data file with module-check records, process inputs and settings, solvents and process designs, background tables (grid, heat and chemical supply, with their factor pack) or credits. The file is read in this browser only and is never uploaded; it replaces the matching records of the active dataset for this session, and every result is recalculated. A file with a section, or a field inside a record, that the page does not know is refused (with the field it was probably meant to be named), so nothing in it is ignored; after loading, a receipt lists what each section applied. The template contains placeholders only: it is rejected until every value is filled in.
File sections
- moduleRecords (with moduleRecordsMode "merge" or "replace-all"): module-check runs, each with its own id and one group: development (fits the module), calibration (sets the transfer bias of the prediction error) or test (scored only). Optional run (the original run id), series (experimental batch or series) and equipmentId, as text. Readings of one run or series must stay in one group, and a run may not appear in two groups with the same conditions and reading: such files are refused. In "merge" a record with the id of a loaded record replaces it.
- processParameters: new ranges (lo, mode, hi) for inputs of a design. Every input must be zero or more, the recovery rate and the usable share of recovered solvent at most 1 and the reboiler heat loss below 1, as in a new design. Only inputs that scale with production (S, rg, el, ht, hr) take a basis; a basis on any other input is refused. A record without basis keeps the basis of the record it replaces, and one without shareKey keeps the shared random draws of the record it replaces ("" clears the key). The usable share of recovered solvent (q) may be added to a design with solvent recovery.
- processSettings: metering scope, failure stage, incinerationShare (share of the solvent waste incinerated, 0 to 1; the rest goes to off-site recovery) and outputForm (the form a design's kg of output refers to, with its activeFraction, above 0 and at most 1). solvents, processes: new solvents (optional fossilC, the fossil share of the solvent's carbon, 0 to 1; a solvent replaced without it keeps its fossil share) and process designs (with the same optional settings).
- grid (kg CO₂-eq per kWh) and heatFuelEF (kg CO₂-eq per MJ of fuel, the boiler input) by region, and chemYear (multiplier of the reference-year chemical footprints, above 0), keyed by year and scenario: "2024", "2033M", "2041L".
- factorPack, needed with any of those three: database, systemModel, method (LCIA method and version), gwpHorizon (100 or 20), biogenicCO2 ("0/0" or "-1/+1"), units (grid "kg CO2-eq/kWh", heatFuelEF "kg CO2-eq/MJ fuel", chemYear "multiplier") and, for future years, scenario (iam, pathway, premiseVersion). A file whose pack differs from one already loaded is refused. Solvent production factors follow a pack too: a solvent file may give one, and must give the loaded pack once a local pack is loaded. A file made for v0.2.0 with background tables has no factor pack and is refused until one is added.
- credits: who refined each module. Other fields allowed: format, version, source, note.
One row for every region: it scales reagents and solvent production. Years not shown are interpolated.
Placeholder values standing in for premise IMAGE SSP2 backgrounds and NESO pathways.
Every heat input (process, solvent recovery, facility) is useful heat; fuel = useful heat ÷ boiler efficiency , then this factor applies.
Methods
How the tool works, module by module. The LCA calculation references PAS 2090:2025 (see Reference standard below). Numbers on this page come from the active dataset and model, so they change when a local data file is loaded.
How a module is reused
Try it: plug modules into a process
Pick a target process, then drag a module card onto its slot (near the top or bottom edge of the window the page scrolls), or double-click a card to place it in its slot. Double-click a slot to take a plugged module out and put the target's own back. With the keyboard or on a phone, select a card (Enter, Space or tap) and then the slot. The result is recalculated with the same draws. Inputs the target lacks are borrowed from another process and listed: that makes the result a what-if, not an assessment.