Perancangan Kontroler Tertanam Menggunakan Direct Neural Network untuk Pengaturan Level Tangki PCT-100

Zuhairi, Muhammad Faris (2023) Perancangan Kontroler Tertanam Menggunakan Direct Neural Network untuk Pengaturan Level Tangki PCT-100. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 07111940000164-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only until 1 October 2025. Download (3MB) | Request a copy

Abstract

Level tangki PCT-100 memiliki karakteristik nonlinear akibat keluaran air drain valve. Kemampuan self learning kontroler dibutuhkan untuk mengatasi perubahan beban. Kontroler direct neural network dapat mengendalikan level dengan mengambil inpur error dan secara langsung mengirim output sinyal kontrol. Namun algoritma cerdas kontroler membutuhkan pemilihan nilai learning rate yang tepat agar mendapat spesifikasi respon yang diinginkan. Spesifikasi ini berupa overshoot level kurang dari 10% dan settling time 5% kurang dari 200 detik saat pembebanan valve dan tracking set point. Selain itu, aktuator flow control valve membutuhkan daya yang besar sehingga harus dibuat kontroler tertanam agar sinyal kontrol langsung melakukan penguatan daya untuk membuka valve. Rangkaian conditioning yang dibuat memberikan hasil linear dengan kesalahan pengukuran maksimal sebesar 121.6mV untuk pembacaan sensor dan 489mV untuk tegangan kontrol. Kontroler neural network dengan learning rate sebesar 10 memiliki hasil terbaik dari sampel uji learning rate konstan, mampu beradaptasi dengan beban motor drain valve dengan overshoot 12% dan settling time 161.7 detik. Adaptasi learning rate melalui parameter alpha juga dapat mengatasi perubahan beban drain terhadap error proses dengan model terbaik saat alpha=0.1, dimana overshoot dan settling time masing-masing sebesar 3.8% dan 111.5 detik untuk simulasi serta 3.43% dan 127.2 detik untuk implementasi. Model neural network yang dibuat mampu melakukan tracking terhadap perubahan set point naik dan turun dengan overshoot dan undershoot yang semakin kecil untuk step set point selanjutnya.
================================================================================================================================
PCT-100 tank level having nonlinear characteristic, caused by valve water outflow. Controller`s learning capabilities is needed to reduce load changes. Direct neural networks controller capable to control process level by taking error as input and directly sending its output as control signal. However, controller`s intelligent algorithm requires selecting learning rate correctly to get desired response specification, that is overshoot less than 10% and settling time 5% less than 200s when load and set point changes. In addition, flow motor valve having high power to be actuated electronically. Embedded controller can be implemented so that control signal amplify motor valve voltage. The conditioning circuit that`s been created provides linear result with a maximum measurement error of 121.6mV for sensor readings and 489mV for control signal. Neural network controller with learning rate=10 achieves the best result among the constant learning rate test samples, capable of adapting water outflow load with 12% overshoot and 161.7 second settling time. Adapting the learning rate through the alpha parameter can overcome changes in the drain load, affecting the process error with alpha=0.1 as the best model with 3.8% overshoot and 111.5 second settling time for simulation, also 3.43% and 127.2 seconds for implementation, respectively. The neural network model is able to track changes in the set point, both increasing and decreasing with diminishing overshoot and undershoot for subsequent set point steps.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	PCT-100, Neural Network, Learning Rate, Sistem Tertanam, Embedded System
Subjects:	Q Science > QA Mathematics > QA336 Artificial Intelligence Q Science > QA Mathematics > QA76.87 Neural networks (Computer Science) T Technology > TC Hydraulic engineering. Ocean engineering > TC424 Water levels T Technology > TK Electrical engineering. Electronics Nuclear engineering > TK3070 Automatic control
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Electrical Engineering > 20201-(S1) Undergraduate Thesis
Depositing User:	Muhammad Faris Zuhairi
Date Deposited:	26 Jul 2023 02:49
Last Modified:	26 Jul 2023 02:49
URI:	http://repository.its.ac.id/id/eprint/98936

Actions (login required)

View Item