Phase-Aware Takagi–Sugeno Fuzzy Predictive Control with Moving-Horizon Estimation for Beer Fermentation

Beer fermentation is a strongly nonlinear and time-varying bioprocess affected by yeast activity, substrate variability, temperature fluctuations, pH changes, and batch-to-batch parameter drift. This study proposes an adaptive predictive control framework that integrates a Takagi–Sugeno fuzzy model, moving-horizon state estimation, and Run-to-Run optimization. A multi-criteria decision-making procedure was initially applied to compare conventional PID, cascade, feedforward, fuzzy-logic, model predictive, and on–off control strategies. The assessment indicated that model predictive control was suitable for the thermally constrained mashing stage, whereas fuzzy predictive control was more appropriate for the nonlinear fermentation process. The fermentation dynamics were divided into lag, active-growth, and stationary phases. A local linear state-space model was constructed for each phase, and the models were combined through Gaussian membership functions to obtain a global Takagi–Sugeno representation. Substrate, biomass, ethanol concentration, and temperature were included in the state vector, while cooling or heating capacity, mixing intensity, flow rate, and pH-adjustment dosage were considered as manipulated variables. A moving-horizon estimator was used to reconstruct unmeasured biological states, and the predictive controller calculated optimal control actions under technological and actuator constraints. Run-to-Run adaptation corrected systematic model bias and compensated for gradual changes between successive fermentation batches. During a 10-hour simulation, the maximum deviation between estimated and reference hidden states remained within ±3%. The probability of incorrect fermentation-phase classification was below 5%. The resulting integral tracking error was 35.7, the normalized control-energy index was 0.048, and no constraint violations were recorded. The control signal remained smooth, reducing excessive actuator activity while maintaining the required substrate-consumption and product-formation trajectories. The proposed framework offers a practical basis for adaptive and energy-efficient beer fermentation control in DMK/SCADA and Industry 4.0 environments.