Fermentation yield-modelling workbench
A bioprocess modelling workbench for fitting fermentation kinetics, comparing operating conditions, and predicting yield from safe prepared or laboratory-generated data.
Project definition
Problem statement
Fermentation yield depends on organism, medium, substrate, temperature, pH, oxygen transfer, mixing, inoculum, and measurement error. A fitted curve can appear accurate while violating mass balance or failing on a new run.
The engineering problem is to compare interpretable kinetic and data-driven models, estimate parameters with uncertainty, and test predictions on runs excluded from fitting.
Project objectives
- Prepare time-series records for selected safe batch or fed-batch fermentation runs.
- Check units, sampling times, missing values, replicate variation, and approximate mass balances.
- Fit selected growth, substrate, and product kinetic models with parameter bounds.
- Compare mechanistic, response-surface, and selected machine-learning predictions.
- Optimise a bounded process objective and test it against held-out or confirmation data.
System design
System modules
Run data manager
Stores organism category, medium, vessel, operating conditions, samples, assays, units, and replicates.
Balance and quality checker
Validates measurements, detects impossible values, and calculates documented carbon or substrate balance proxies.
Kinetic model library
Implements selected growth, substrate-consumption, product-formation, inhibition, and fed-batch equations.
Parameter estimator
Fits bounded parameters, compares residuals, estimates uncertainty, and flags unidentifiable models.
Optimisation dashboard
Compares runs and models and explores bounded yield, productivity, time, and operating-condition tradeoffs.
Methodology
System workflow
- 01Load fermentation runs
The student selects prepared safe datasets and verifies conditions, measurements, units, and replicates.
- 02Check data
Quality rules and approximate balances identify exclusions and uncertain measurements.
- 03Fit models
Selected kinetic and data-driven models use training runs with saved bounds and initial values.
- 04Validate
Predictions are compared with held-out runs, residual patterns, and parameter uncertainty.
- 05Explore conditions
A bounded optimiser proposes a tradeoff and the system compares it with confirmation data where available.
Demonstration scenario
Several prepared yeast fermentation runs are used to fit Monod-type growth and product models. One run remains held out. The workbench compares measured and predicted biomass, substrate, and product, shows parameter uncertainty and residuals, and tests a bounded temperature and feed scenario.
Engineering
Technical architecture
- Web application
- Nuxt and Vue for run entry, process diagrams, curves, residuals, parameter tables, and optimisation.
- Modelling API
- FastAPI for datasets, equations, fit jobs, uncertainty, validation, optimisation, and exports.
- Data layer
- PostgreSQL for runs, conditions, samples, measurements, model versions, parameters, and predictions.
- Numerical engine
- Python, NumPy, and SciPy for differential-equation integration, constrained fitting, sensitivity, and optimisation.
- Validation design
- Replicate analysis, run-level holdouts, residual diagnostics, parameter identifiability, and balance checks.
Testing
Evaluation
Evaluation measures
- Root mean squared and normalised prediction error for biomass, substrate, and product
- Parameter confidence intervals, correlation, and identifiability
- Approximate mass-balance closure under documented assumptions
- Performance on fermentation runs excluded from fitting
- Sensitivity to starting values, bounds, sampling frequency, and measurement noise
- Integration, fitting, and optimisation time across model and run sizes
System boundaries
- The project uses public, synthetic, or supervised laboratory data from non-pathogenic teaching organisms.
- It does not provide protocols for culturing pathogens, toxins, or hazardous biological agents.
- Model recommendations remain within documented experimental ranges and require laboratory confirmation.
- Biological work must follow institutional biosafety, waste, equipment, and supervision requirements.
Included
- 01Fermentation run and measurement interface
- 02Mass-balance, kinetic, parameter-fitting, and optimisation modules
- 03Growth, substrate, product, residual, and uncertainty dashboards
- 04Prepared non-pathogenic fermentation datasets and results
- 05Complete source code in a private GitHub repository
- 06Complete project documentation with synopsis, abstract, methodology, bioprocess diagrams, modelling results, screenshots, and conclusion
- 07Setup and usage guide
Project record
No buyer information is collected on this page.
- Permanent project ID
- GP-BT-0Y5JQKG
- Catalogued
- 21 Aug 2026
- Completed
- Pending
- Verified
- Pending
- Demonstration
- Added when ready
Handover
After purchase
- 01Payment is confirmed
The project is marked unavailable and cannot be purchased again.
- 02Repository access is granted
The buyer's submitted GitHub account receives access to the private repository.
- 03The purchase record is delivered
The certification sheet is prepared from the reviewed buyer details and sent privately by email.