| Abstract [eng] |
Project‘s goal is to optimize fed-batch biotechnological process using radial basis function network. The work examines modeling of mass balance based biotechnological processes, optimization of technological parameters, structure, application and operation of artificial neural networks. Paper on fed-batch biotechnological process is analyzed. The work consists of fed-batch process imitation model where Pontriagin‘s principle of maximum is used to calculate optimal feeding profile. Later optimal feeding profile is substituted with approximation of radial basis function neural network. Neural network is optimized in six different cases, using evolutionary optimization method. Methods goal function is to maximize the ammount of product (protein) at the end of cultivation process. In all of the cases, state variables x, s and px deviations of trajectories in process ending points, in comparison with reference data, are less than 10% which is the accuracy of measurements in real biotechnological process of biomass, substrate concentration and specific protein activity. Research results show that using parametric optimization method and radial basis function network, optimization results are analogical to reference data. This confirms that results obtained from both methodologies match experimental data and are correct. |
