00000ctm a22000003a 4500 UP-8027390931314102126 Buklod 20240320091938.0 m |o d | ta 240320s xx d r |||| u| (iLib)UPVIS-00163083906 gift VML/ess 03.20.2024 eng LG 993.5 2022 E62 C66 Consular, Mark Rudan Artificial neural network based surrogate modelling for optimization of an industrial batch fermentation process. 2022. xiv, 105 leaves. text rdacontent unmmediated rdamedia volume rdacarrier Thesis (Undergraduate, Bachelor of Science in Chemical Engineering) School of Technology, University of the Philippines Visayas, Miagao, Iloilo. 2022. Fermentation is a well-known process to produce alcohol. Optimization of this process reduces the production costs while maximizing the ethanol yield. This study aims to model the batch fermentation process of Asian Alcohol Corp. and optimize it through a surrogate model that predicts the potential profit based on artificial neural network using MATLAB R2021b software. Andrews model was used and simulated applying the fourth order Runge-Kutta method ODE15s. The kinetic parameters were estimated using fminsearch function. Generation and training of the surrogate model were done using fitrnet function augmented with 5-fold cross validation OptimizeHyperparameters. Predicted potential profit was then optimized through particle swarm optimization. The estimated parameters of the kinetic model were μmax=2.5519 h-1, KS=25.8553 g.L-1, Ki=187.9845 g.L-1, kd=0.61943 h-1, YX/S=1.5668 ×10-4, and YP/S=0.6757. The surrogate model has an elapsed time of 8.367 hrs, layer size of [176 64 66], estimated objective function of 21.1334, regularization parameter (λ) = 3.185×10-2 and relu as activation function (σ). This model has projected an R2=0.99993 and RMSE = 15,555.987. The optimal conditions of 3.27 ×107 CFU/mL initial cell biomass, 111.74 MT initial molasses and 33.36 hours of fermentation time were predicted to give a potential profit of Php 1,066,440.59. Using sensitivity analysis, the influential parameters of the process are the amount of initial molasses used and the fermentation time. Percent error ranging from 29.3795% to 33.6129% was obtained between the industrial and estimated potential profit by the surrogate model. Artificial neural network. Calangi, Geraldine D. Thesis adviser. FI UPVIS UPV-SOTECH LG 993.5 2022 E62 C66 Thesis