PENGGUNAAN JARINGAN SYARAF TIRUAN UMPAN MAJU DAN DIAGONAL RECURRENT UNTUK IDENTIFIKASI DAN KONTROL SISTEM DINAMIK

• Main Author: Hafid , Abdul
• Format: Theses
• Language: Indonesia
• Subjects: Teknik (Rekayasa, enjinering dan kegiatan berkaitan)
• Online Access: https://digilib.itb.ac.id/gdl/view/4866
• Institution: Institut Teknologi Bandung

<b>Abstract :</b><p align=\"justify\"> <br /> During the last few years, a technique which has been investigated by many researcher and applied in many diverse areas such as signal processing, pattern recognition, identification and control system, is Artificial Neural Networks (ANN). In particular, for the purpose of systems identification and control of dynamical systems, the architectures of ANN, which are mostly applied, are Feedforward and Diagonal Recurrent Neural Networks (DRNN). In the process of identification and control of a plant using ANN, its necessary to conduct learning process. The learning algorithm which is commonly used, is Backpropagation (BP). Other alternative is to apply Recursive Prediction Error (RPE). <br /> <p align=\"justify\"> <br /> This thesis discusses three main topics in connention with identification and control using artificial neural networks. Firstly, how to improve the performance of RPE algorithm by introducing a factorization technique of matrix P in the algorithm. Next discussing deals with modelling, i.e. to develop a mathematical model of a plant based-on the result of identification using ANN. In the modelling process, the performance of four different structures of ANN, developed in this investigation, is compared. Last topic discusses a design of control system based-on DRNN. This control method can be applied effectively provided that its learning rate is adjusted adaptively. The identification procedure and the designed control method is then applied for modelling and control of vechile suspension system and turbogenerator plant, using input-output data obtained from real time experiments. <br /> <p align=\"justify\"> <br /> Observation show that the designed ANN techniques perform well in identification of the plant and produced mathematical model of the plant. Good control performance was also obtained in the case of controlling the suspension deflection of the vechile and turbine rotating frequency of turbogenerator plant .