DESAIN DAN ANALISA PENENTUAN PARAMETER GAIN SISTEM PENGENDALI PROPORTIONAL-INTEGRAL DENGAN METODE NEURAL NETWORK PADA SISTEM TURRET GUN

NOOR, DERIS TRIANA (2017) DESAIN DAN ANALISA PENENTUAN PARAMETER GAIN SISTEM PENGENDALI PROPORTIONAL-INTEGRAL DENGAN METODE NEURAL NETWORK PADA SISTEM TURRET GUN. Undergraduate thesis, Institut Teknologi Sepuluh Nopember.

[img]
Preview
Text
2112100115-Undergraduate_Theses.pdf - Published Version

Download (3MB) | Preview

Abstract

Parameter dalam sistem pengendali PID untuk sistem turret gun ditentukan dengan cara menurunkan persamaan diferensial sistem pengendali tersebut. Penurunan persamaan diferensial memakan banyak waktu dan sulit untuk dilakukan. Terutama ketika sistem yang dikontrol sangat rumit dan sulit untuk disederhanakan, sehingga sistem tidak dapat diwakili oleh suatu persamaan. Untuk menanggulangi masalah tersebut, digunakanlah neural network untuk menentukan parameter sistem pengendali. Sistem pengendali yang digunakan adalah sistem pengendali proportional-integral yang hanya memerlukan nilai proportional gain (KP) dan integral gain (KI). Nilai KP dan KI digunakan sebagai faktor pengali pada nilai error yang menjadi input dari sistem pengendali proportional-integral dan input neural network. Dengan demikian, nilai dari KP dan KI pada sistem pengendali proportional-integral diperoleh melalui pelatihan neural network. Dari Tugas Akhir ini didapatkan hasil nilai KP dan KI yang dapat berubah setiap waktu sesuai error yang terjadi pada sistem. Pada uji error 10 sudut yang dipilih secara acak, didapatkan bahwa nilai sudut aktual dari sudut azimuth dan sudut elevasi sudah mendekati nilai inputnya dengan nilai root mean square error (RMSE) adalah 0.0393063 untuk sudut azimuth dan 1.1621708 untuk sudut elevasi. Respon sistem turret gun dengan Neural Network Proportional-Integral (NN-PI) controller dapat menurunkan %overshoot 63,493%, settling time 19,183%, dan steady-state error 27,064%. Sementara untuk sudut elevasi, Neural Network Proportional-Integral controller hanya dapat menurunkan steady-state error yaitu 34,168 %. Sedangkan untuk settling time mengalami kenaikan 16,463% dan %overshoot juga mengalami kenaikan 75,465%. =========================================================================================== Parameters in PID controller for turret-gun system are determined by derivation of the differential function. The derivation might be difficult and take many times to be solved, especially if the systems are complex and difficult to obtain the exact function which has the same behavior as the system. To slove the problem, neural network are proposed to determine the parameters of the controller. The controller which applied in this project was Proportional-Integral (PI) controller. It has two parameters which was proportional gain (KP) and integral gain (KI). Both were used as gain factor for the error value which at the same time have a role as an input value to the controller and the neural network. So, the gain value for the PI controller determined by the neural network training according to the error value. The result of this final project was obtain the parameter gain value that has the ability to adapt in every condition according to error value that happens in the system. From the error value test, ten set points were choosed randomly. The result of this test was the azimuth angle and elevation angle have the ability to keep up with the set point. Despite the ability of the controller, there was still an error in every angle. The root mean square error (RMSE) was 0.0393063 for the azimuth angle and 1.1621708 for elevation angle. Response of the system with neural network proportional-integral (NN-PI) controller for azimuth angle have the ability to decrease the %overshoot value to 63,493 %, settling time value to 19,183 %, and the steady state error value to 27,064 %. While for the elevation angle, the controller only have the ability to decrease the steady state error value to 34,168 %. But for settling time value has increased to 16,463 % and %overshoot value to 75,465 %.

Item Type: Thesis (Undergraduate)
Uncontrolled Keywords: turret gun, neural network, proportional-integral control, azimuth, elevasi.
Subjects: T Technology > TJ Mechanical engineering and machinery
T Technology > TJ Mechanical engineering and machinery > TJ217.6 Predictive Control
Divisions: Faculty of Industrial Technology > Mechanical Engineering > 21201-(S1) Undergraduate Thesis
Depositing User: DERIS TRIANA NOOR
Date Deposited: 25 Jan 2017 06:22
Last Modified: 06 Mar 2019 07:06
URI: https://repository.its.ac.id/id/eprint/3190

Actions (login required)

View Item View Item