Answers are grounded in this paper and cite their sources.
General Dynamical Model (GDM)-based observers estimate unknown bioprocess kinetics, but their reconstruction maps vary across strains. We formulate a minimax design that jointly selects measurement configuration, nominal model and observer tuning using a population-structured uncertainty set and an analytical reconstruction-bound objective. Evaluation uses averaged hourly trajectories from 17 Kluyveromyces marxianus strains and model-derived pathway-rate references. At the prescribed 5% disturbance scenario, the selected substrate-product (SP) design has a bound of 1.4045, 39.5% below biomass-product (XP). Strict training-only leave-one-strain-out validation selects SP in all 17 folds. Eleven folds pass the implemented structural, rate-envelope and kinetic-variation screens, and their exact nodal-box errors remain below the reported asymptotic benchmark and finite-time bounds. These numerical screens do not independently verify continuous-time model adequacy or intersample assumptions. At 5%, SP has the lowest exact error in 15/17 folds under linear interpolation. Empirical reconstruction accuracy depends on the disturbance level and propagation rule; the bound-selected SP design does not consistently outperform the zero-estimator baseline, particularly under causal zero-order hold (ZOH) and at the prescribed 10% disturbance level. The contribution is a bound-based design formulation with conditional analytical guarantees; empirical accuracy depends on the pathway, disturbance scenario and implementation.
No takes yet. Share an insight, caveat, or question.
Lyubenova et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: