Deteksi Kerusakan Pompa Berdasarkan Sinyal Vibrasi Menggunakan Machine Learning.

Ihsannur, Haris (2022) Deteksi Kerusakan Pompa Berdasarkan Sinyal Vibrasi Menggunakan Machine Learning. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 02311840000086-undergraduate_thesis.pdf Restricted to Repository staff only Download (6MB)

Abstract

Di dunia industri, monitoring kondisi sangat dibutuhkan karena dapat meningkatkan keandalan dari suatu komponen mesin. Apabila terjadi kerusakan dan tidak segera diperbaiki, maka akan menimbulkan kerusakan yang lebih parah. Jenis-jenis kerusakan tersebut menghasilkan sinyal vibrasi dengan pola tertentu. Tujuan dari penelitian ini yaitu dapat mendeteksi kerusakan mesin secara otomatis menggunakan machine learning. Dataset berupa sinyal getaran pompa dalam domain waktu diubah dalam domain frekuensi menggunakan teknik Fast Fourier Transform (FFT). Hasil FFT tersebut kemudian diekstrak fiturnya. Terdapat 9 fitur ekstraksi statistik yang digunakan yang kemudian dipilih fitur terbaik menggunakan teknik feature importance score. Hasil ekstraksi dimasukkan ke dalam algoritma machine learning. Terdapat 3 algortima machine learning yaitu k-nearest neighbor (kNN), support vector machine (SVM), dan gaussian naive bayes (GNB). Melalui proses training dan validasi sehingga menghasilkan nilai unweighted accuracy (UA) serta dapat memprediksi jenis kerusakan pada pompa. Nilai UA dengan algoritma kNN, SVM, dan GNB yaitu 95,80%, 95,29%. dan 80,69%. Pengujian algoritma dilakukan dengan program inference yang bertujuan untuk mengevaluasi algoritma machine learning yang dirancang.
===================================================================================================================================
In the industrial world, condition monitoring is needed because it can increase the reliability of a machine component. If damage occurs and is not repaired immediately, it will cause more severe damage. These types of damage produce a vibration signal with a certain pattern. The purpose of this research is to detect machine breakdowns automatically using machine learning. The dataset in the form of a pump vibration signal in the time domain is converted into the frequency domain using the Fast Fourier Transform (FFT) technique. The results of the FFT are then extracted features. There are 9 statistical extraction features used which are then selected the best features using the feature importance score technique. The extraction results are entered into a machine learning algorithm. There are 3 machine learning algorithms, namely k-nearest neighbor (kNN), support vector machine (SVM), and gaussian naive bayes (GNB). Through the process of training and validation to produce an unweighted accuracy (UA) value and can predict the type of damage to the pump. The UA values with the kNN, SVM, and NAM algorithms are 95.80%, 95,29%, and 80.69%. Algorithm testing is done with an inference program that aims to evaluate the designed machine learning algorithm.

Item Type:	Thesis (Other)
Additional Information:	RSF 621.67 Ihs d-1 2022
Uncontrolled Keywords:	vibrasi, FFT, machine learning. vibration, FFT, machine learning
Subjects:	T Technology > TJ Mechanical engineering and machinery > TJ919 Centrifugal pumps--Design and construction.
Divisions:	Faculty of Industrial Technology and Systems Engineering (INDSYS) > Physics Engineering > 30201-(S1) Undergraduate Thesis
Depositing User:	Mr. Marsudiyana -
Date Deposited:	11 May 2026 08:10
Last Modified:	11 May 2026 08:10
URI:	http://repository.its.ac.id/id/eprint/133132

Actions (login required)

View Item