Adaptive Digital Twin
Research & roadmap
Modica is an R&D-stage company building an adaptive digital twin for bioprocesses. Here is where we are, the thesis behind the work, the problems that make it genuinely hard, and where the platform — and the services around it — are heading.
Where we are
Active R&D on the adaptive digital twin: hybrid process modeling, live prediction with quantified uncertainty, and a universal connectivity layer that standardizes plant data. Co-financed by the EU under the FENG programme.
About the EU-funded projectOur research thesis
A model that only fits history is a report; a model that generalises — to the next batch, the next scale, the next molecule — is a tool. We believe that generalisation comes from combining mechanism with data: a mechanistic backbone that encodes what we know about growth, mass transfer and metabolism, and machine learning for what we do not. That keeps the twin interpretable and lets it become useful on far fewer runs than a purely data-driven model would need.
The problems that make this hard
Adaptivity across scale
A model calibrated at 20 L should carry over to 2,000 L without starting from scratch. We treat scale-up and tech transfer as a transfer-learning problem, not a rebuild.
Honest uncertainty
A prediction without a confidence band invites false confidence. The twin has to say how sure it is — and be right about being unsure — so an operator knows when to trust it and when to look closer.
Living calibration
Biology drifts. A twin that is faithful today can be wrong next campaign. Keeping the model in step with the process automatically is a core problem, not a footnote — it is what the “adaptive” in adaptive digital twin has to earn.
What Modica can provide
Digital-twin modeling
Building and calibrating process models for a specific bioprocess, with golden-envelope monitoring and prediction.
Connectivity & integration
Standardizing plant data across OPC UA, Modbus, S7, MQTT and MTP — and accelerating connector and configuration build-out.
Decision support
Turning forecasts into prescriptive recommendations for operators and process engineers.
Where we are heading
- Broader equipment and protocol coverage in the connectivity layer
- Richer prescriptive control and scenario simulation
- Centralized analytics across parallel batches and sites