Digital twins

A digital twin is a model that mirrors a physical system so you can understand it, calibrate it and anticipate its behavior before acting on the real equipment. I work on this idea applied to bioreactor operation: vessels where microorganisms are grown to produce proteins, biofuels or metabolites.

A concrete case is the pipeline from the MSc thesis of Gonzalo I. Barrera Hernández (which I co-advised): engineered E. coli produces an orange fluorescent protein (OFP) from glucose. The challenge has two parts —and so it is solved in two stages chained with evolutionary algorithms.

Explore it

This digital twin runs the full pipeline in your browser; it is not an animation:

• Stage ① · Batch (calibration). We don't know the organism's kinetic parameters. A multi-objective evolutionary algorithm (NSGA-III style) tunes them so the model reproduces real lab data across four objectives at once —biomass, substrate, product and oxygen. You'll watch the R² values climb toward 1.
• Stage ② · Fed-batch (optimization). With a faithful model, a genetic algorithm decides how to feed the reactor (S_in) to maximize biomass without running out of oxygen (≥ 20%) or exceeding the volume (≤ 2 L).

It is a didactic model, not a twin calibrated to a specific vessel, but the pipeline —calibrate with an EA, then optimize with a GA— is exactly the thesis's. It ties into my other lines: the multiobjective optimization that powers stage ①.