Rancang Bangun Speed Sensorless Fault Tolerant Control Pada Motor DC Menggunakan Machine Learning Classifier

Setiawan, Ahmad Bagas (2023) Rancang Bangun Speed Sensorless Fault Tolerant Control Pada Motor DC Menggunakan Machine Learning Classifier. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 02311940000049-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only until 1 September 2025. Download (3MB) | Request a copy

Abstract

Pada beberapa tahun belakangan pengendalian motor DC telah berkembang sebuah sistem speed sensorless motor DC yaitu sebuah proses pengendalian kecepatan motor tetapi tidak menggunakan sensor kecepatan. Speed sensorless motor DC dinilai lebih efektif. Pada prosesnya sistem speed sensorless akan menggunakan observer sebagai esimasi kecepatan. Sedangkan dalam sistem kontrol kecepatan rentan untuk mengalami kesalahan seperti perubahan torsi beban dan kesalahan sensor. Dikarenakan hal tersebut diterapkan fault tolerant control (FTC) pada sistem kontrol untuk mengatasi kesalahan tersebut. Dikarenakan pada permasalahan ini terdapat dua jenis kesaahan maka diperlukan fault detection and isolation (FDI) untuk membedakan kedua jenis kesalahan tersebut. Pada penelitian ini FDI yang digunakan adalah machine learning claasifier untuk melakukan pembelajaran pintar sebagai sistem deteksi kesalahan. Pada penelitian ini didapatkan hasil bahwa penerapan machine learnig classifier decision tree dapat mengidentifikasi secara baik apakah dalam kondisi healthy, perubahan torsi beban, atau kesalahan sensor arus dengan nilai akurasi 99.926%, presisi sebesar 99.889%, dan nilai recall sebesar 99.889%. Integrasi classifier dengan sistem kontrol FTC dapat bekerja dengan cukup baik dengan hasil ketika diberikan perubahan torsi beban memiliki setling time sekitar 3.5 detik, overshoot setelah diberikan perubahan torsi beban sebesar 4.7% sedangkan setelah diberikan kesalahan sensor sebesar 10%, nilai error steady state setelah diberikan perubahan torsi beban sebesar 2% sedangkan setelah diberikan kesalahan sensor sebesar 3.5%.
=================================================================================================================================
In recent years, DC motor control has developed a DC motor speed sensorless system, which is a process of controlling motor speed but not using a speed sensor. Speed sensorless DC motors are considered more effective. In the process, the speed sensorless system will use an observer as a speed estimate. Meanwhile, the speed control system is prone to errors such as changes in load torque and sensor errors. Because of this, fault tolerant control (FTC) is applied to the control system to overcome these errors. Because there are two types of errors in this problem, fault detection and isolation (FDI) is needed to distinguish the two types of errors. In this study, the FDI used is a machine learning classifier to conduct smart learning as an error detection system. In this study, the results showed that the application of the machine learning classifier decision tree can identify well whether in healthy conditions, changes in load torque, or current sensor errors with an accuracy value of 99.926%, a precision of 99.889%, and a recall value of 99.889%. The integration of the classifier with the FTC control system can work quite well with the result that when given a load error it has a settling time of about 3.5 seconds, overshoot after being given a load error of 4.7% whereas after being given a sensor error of 10%, the steady state error value after being given a load error of 2% while after being given a sensor error of 3.5%

Item Type:	Thesis (Other)
Uncontrolled Keywords:	DC Motor, Fault Tolerant Control, Machine Learning Classifier, Motor DC, Observer, Speed Sensorless
Subjects:	T Technology > TL Motor vehicles. Aeronautics. Astronautics > TL521.3 Automatic Control
Divisions:	Faculty of Industrial Technology and Systems Engineering (INDSYS) > Physics Engineering > 30201-(S1) Undergraduate Thesis
Depositing User:	Ahmad Bagas Setiawan
Date Deposited:	02 Aug 2023 08:12
Last Modified:	02 Aug 2023 08:12
URI:	http://repository.its.ac.id/id/eprint/100322

Actions (login required)

View Item