Fault Detection Pada Pompa Sentrifugal Menggunakan Vibration Analysis Berbasis Machine Learning

Wicaksono, Indrajit Aryo (2023) Fault Detection Pada Pompa Sentrifugal Menggunakan Vibration Analysis Berbasis Machine Learning. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 04211940000079_Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only until 1 September 2025. Download (23MB) | Request a copy

Abstract

Predictive maintenance dilakukan dengan memprediksi kondisi aset melalui condition monitoring. Namun, perusahaan harus dapat menyimpulkan kondisi aset dari data yang dimiliki. Studi ini membahas masalah prediksi kegagalan pompa sentrifugal menggunakan model machine learning yang dilatih menggunakan data getaran. accelerometer digunakan untuk mengukur getaran aksial, horizontal, dan vertikal dari pompa yang telah dikondisikan sebelumnya dalam empat kondisi: unbalance, misalignment, normal, dan bearing defect. Data yang diperoleh kemudian dibersihkan atau diubah melalui Fast Fourier Transform (FFT) dan digunakan untuk melatih algoritma klasifikasi pada machine learning seperti decision tree, random forest, K nearest neighbour, dan discriminant analysis. Model machine learning yang dibuat memungkinkan prediksi kondisi pompa. Akurasi model dipengaruhi oleh faktor-faktor seperti jumlah fitur dimana semakin banyak pengukuran getaran yang digunakan maka akurasi model juga semakin tinggi, pilihan algoritma menunjukkan bahwa hanya algoritma discriminant analysis yang menunjukkan ketidakmampuan untuk memprediksi pada level algoritma lain ketika menggunakan data frequency domain karena data tidak terdistribusi secara normal, metode validasi dapat meningkatkan akurasi model namun dengan waktu pelatihan yang lebih lama, dan jenis data yang digunakan menunjukkan bahwa data frequency domain dapat menjelaskan gangguan dengan lebih baik dibandingkan dengan data time domain. Temuan menunjukkan penerapan
=================================================================================================================================
Predictive maintenance is done by predicting the condition of asset through condition monitoring of asset. However, companies must be able to infer from the data the condition of the asset. This study tackles the issue of centrifugal pump failure prediction using machine learning model trained using vibration data. Accelerometers were used to measure the Axial, horizontal, and vertical vibrations of a pre-conditioned pump under four conditions: imbalance, misalignment, normal operation, and bearing problem. Data acquired then cleaned or transformed through Fast Fourier Transform (FFT) and were used to train machine learning classification algorithms such as decision tree, random forest, K nearest neighbour, and discriminant analysis. Model created learning enabling the prediction of pump conditions. Models accuracy was found to be influenced by factors such as the amount of feature where the more vibration measurement used the accuracy of the model also become higher , the choice of algorithm shows that only discriminant analysis algorithm that shows inability to predict to the level of other algorithm when using frequency domain data due to data not normally distributed, the validation method can improve model accuracy but it comes with longer training period, and the type of data used showed that frequency domain data can explain the fault better than time domain data. The findings demonstrate the application of machine learning in predicting pump failure using vibration data.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Algorithm, Centrifugal Pump, Fast Fourier Transform, Machine Learning, Predictive Maintenance, Vibration Analysis,Algoritma, Analisis Getaran, Fast Fourier Transform, Machine Learning, Predictive Maintenance, Pompa Sentrifugal.
Subjects:	T Technology > TJ Mechanical engineering and machinery > TJ174 Maintenance and repair of machinery
Divisions:	Faculty of Marine Technology (MARTECH) > Marine Engineering > 36202-(S1) Undergraduate Thesis
Depositing User:	Indrajit Aryo Wicaksono
Date Deposited:	07 Aug 2023 02:09
Last Modified:	07 Aug 2023 02:09
URI:	http://repository.its.ac.id/id/eprint/101783

Actions (login required)

View Item