WP1 studies how manufacturing models transfer between medicines and formulations. This engineering prototype tests one component: an electricity model for aqueous spray drying on the same equipment and meter boundary, conditional on declared target inputs.
What is fixed and what is supplied
The source evaporation coefficient α, auxiliary power, and fixed electricity remain fixed. Target collection yield, residual moisture, and duration are supplied before locking the prediction. They determine the predicted dry output and electricity intensity. Product quality is checked from observations after locking.
Mass balance
Dry output = feed mass × solids mass fraction × collection yield
Evaporated water = feed mass - feed mass × solids mass fraction ÷ (1 - residual moisture fraction)
Residual moisture is expressed on a wet basis. Collected product and lost powder have the same moisture content, and collection loss occurs after drying. The mass balance includes water vapor and powder as its outputs and excludes liquid discharge, solvent recirculation, and chemical reaction. The output denominator is dry formulation solids. Calculating energy per kilogram of API also requires the API content.
Electricity model
Source coefficient α = (source batch electricity - auxiliary power × source duration - fixed electricity) ÷ source evaporated water
Target batch electricity = α × target evaporated water + auxiliary power × target duration + fixed electricity
The fitted coefficient α has units of kWh/kg of evaporated water. Auxiliary power and fixed electricity are supplied separately, so the source batch total fits only α. Setting both terms to zero assigns all source electricity to evaporation. Source and target use matching meter boundaries and duration definitions. Inlet temperature is checked against the declared range.
Acceptance metric and baseline
The primary electricity metric is kWh per kg of collected dry formulation solids. Its signed error is (predicted intensity - observed intensity) / observed intensity. A pass requires this intensity error and the dry-output deviation to meet their separate tolerances, measured quality to meet its criterion, and the independence, declared-range, and batch-boundary checks to pass. Batch-total electricity error is reported separately as a diagnostic and has no pass threshold.
The baseline multiplies source electricity intensity by predicted target dry output. Both models are compared against observed electricity intensity using the observed dry output. The outcome applies to the tested target batch and its locked input assumptions.
How this supports full WP1
This version records medicine or formulation identity, locks conditional predictions, checks observations, and compares errors with a direct-scaling baseline. The researcher declares the input ranges before viewing outcomes. These bounds are an eligibility check. Empirical transfer boundaries require tests across several target medicines and formulations.
The full WP1 study will develop learned quality models, compare fixed transfer with limited target calibration, test how input shifts relate to prediction error, and evaluate prediction intervals. These analyses require additional model modules and independent target data.
Locked records
A lock saves a complete snapshot, timestamp, and SHA-256 fingerprint. Later edits create a new working version while the original snapshot remains available for export. Imported snapshots are checked for changes. Local locks document the workflow. Independent preregistration requires depositing the snapshot with an external registry.
Preparing research data
Converting electricity per unit of output to batch electricity requires the corresponding output quantity and normalization basis. Repeated operating conditions need traceable batch identifiers. Electricity measured with water as a substitute for another solvent must be recorded as a substitute-medium measurement. Check the original units and evidence before fitting a source record.
Public equipment and experimental records
The equipment library lists manufacturer ratings separately from published observations. Selecting a record supplies its reported feed rate and concentration for dry-solids input arithmetic. Optional total feed mass gives feed-only duration, with startup and cleaning outside that calculation. Batch electricity, product yield and quality predictions require additional measurements and models.
Source drafts copy only verified source inputs into an otherwise empty research project. Public outcomes are known development evidence. The physical equipment ID, meter boundary, energy parameters, quality limits and target test conditions require independent documentation.