#TITLE_ALTERNATIVE#

• Main Author: Merry Sartika<br>NIM. 23398007, Erwani
• Format: Theses
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/6074
• Institution: Institut Teknologi Bandung

<b></i>Abstract:</b><i><p align="justify"><br />Fermentation process is a very well known process in the human history specifically in producing food such as tempe, yoghurt, and cheese, etc. However, until now there is still lack of in-depth research about how to control the result of the process. One thing that makes the control of the fermentation is not developed properly is because this process has a complicated characteristic which involves many process variables, physically and chemistry variables.<br /><p align="justify"><br />An approach to solve this problem is by using Iterative Learning Control (ILC). ILC is a technique to improve the performance of a system or a process which is operated either in the batch, fedbatch, or continuous mode. ILC will provide significant improvement when the system is subjected to several kinds of inputs and model uncertainties in both design and model. ILC is a design of a controller which will ensure a desired trajectory with an acceptable accuracy.<br /><p align="justify"><br />The aim of this thesis is to design a controller for a fed batch process which enables the growth of yeast known as Saccharomyces cerevisiae with the ILC method. Based on the growth profile determined, ILC method is used to get a progress pattern of glucose which enables the growth of Saccharomyces cerevisiae like a growth profile determined.<br /><p align="justify"><br />In particular, in this research, a control is designed for a fed-batch fermentation process by using two types of ILC methods, i.e. Finite Horizon Learning Control (FHLC) and Non Linear Learning Control (NLC). In FHLC, it is assumed that the system is linear, in which case, a stair-case function input with a tight progress is needed to get a better result. For example for a stair-ease with 8 stairs resulted in integral error 0.05. On the other hand, NLC is undertaken by using 3 kinds of stream velocity in the form of sinusoid, staircase and step. There are 5 to 8 learning iteration needed by NLC for sinusoid with integral error 0.0016 and staircase with integral error 0.003; for step there are 20 iteration needed with integral error 0.0111.<br />