Perancangan Sistem Deteksi Kesalahan Berbasis Neural Network - Residual Analysis Pada Water Injection Pump Di PT. Saka Indonesia Pangkah Limited

Yugoputra, Gabriel Fransisco (2021) Perancangan Sistem Deteksi Kesalahan Berbasis Neural Network - Residual Analysis Pada Water Injection Pump Di PT. Saka Indonesia Pangkah Limited. Undergraduate thesis, Institut Teknologi Sepuluh Nopember.

Text 02311740000039-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only until 1 October 2023. Download (2MB) | Request a copy

Abstract

Pompa sentrifugal merupakan alat yang populer dan sangat sering digunakan untuk memindahkan cairan di industri. Sistem deteksi dan diagnosis kesalahan menjadi praktik yang sangat penting untuk mencegah kerusakan yang dapat mengganggu kinerja pompa. Deteksi dini kesalahan dapat membantu menghindari kerusakan produk, penurunan kinerja, kerusakan pada mesin, dan kerusakan pada kesehatan manusia, atau bahkan hilangnya nyawa. Pada penelitian ini, dirancang sistem deteksi kesalahan pada Water Injection Pump berbasis Neural Network – Residual Analysis. Residual merupakan selisih antara sinyal variabel proses pompa dan sinyal model jaringan syaraf tiruan. Data didapat dari PT. Saka Indonesia Pangkah Limited yang terdiri dari data historis pompa yang meliputi kondisi healthy serta faulty. Kesalahan yang dianalisis diantaranya suction strainer kotor dan kesalahan pressure transmitter. Data variabel yang digunakan dalam penelitian ini adalah arus, flowrate, dan differential pressure. Pemodelan jaringan syaraf tiruan yang diterapkan menggunakan arsitektur feed-forward backpropagation dengan algoritma pelatihan Levenberg-Marquardt. Kemudian dilakukan ekstraksi fitur residual untuk mendapatkan parameter statistik dari residual yang dihasilkan. Classifier dirancang untuk mengklasifikasi kondisi pompa berdasarkan data parameter statistik dari tahap ekstraksi fitur. Adapun akurasi classifier yang telah dibuat sebesar 94,8% untuk Gaussian Naïve Bayes dan 93,8% untuk k-Nearest Neighbor.
======================================================================================================
Centrifugal pumps are a popular tool and are very often used to move fluids in industry. Fault detection and diagnosis systems are becoming a very important practice to prevent malfunctions that can impair pump performance. Early detection of faults can help avoid product breakdowns, performance degradation, damage to machines, and damage to human health, or even loss of life. In this research, an fault detection system for Water Injection Pump based on Neural Network - Residual Analysis was designed. Residual is the difference between the pump process variable signal and the neural network model signal. Data obtained from PT. Saka Indonesia Pangkah Limited which consists of historical pump data covering healthy and faulty conditions. The fault analyzed include dirty suction strainer and pressure transmitter fault. The data variables used in this study are current, flowrate, and differential pressure. The artificial neural network modeling is applied using a feed-forward backpropagation architecture with the Levenberg-Marquardt training algorithm. Then the residual feature extraction is carried out to obtain statistical parameters from the resulting residuals. The classifier is designed to classify pump conditions based on statistical parameter data from the feature extraction stage. The accuracy of the classifier that has been made is 94.8% for Gaussian Naïve Bayes and 93.8% for k-Nearest Neighbor.

Item Type:	Thesis (Undergraduate)
Uncontrolled Keywords:	Classifier, Deteksi Kesalahan, Fitur Parameter Statistik, Levenberg-Marquardt, Neural Network, Residual Analysis, Water Injection Pump, Fault Detection, Statistical Parameter Feature
Subjects:	Q Science > Q Science (General) > Q325.5 Machine learning. Q Science > QA Mathematics > QA336 Artificial Intelligence Q Science > QA Mathematics > QA76.87 Neural networks (Computer Science) T Technology > T Technology (General) > T57.5 Data Processing T Technology > TJ Mechanical engineering and machinery > TJ903 Pumping machinery.
Divisions:	Faculty of Industrial Technology and Systems Engineering (INDSYS) > Physics Engineering > 30201-(S1) Undergraduate Thesis
Depositing User:	Gabriel Fransisco Yugoputra
Date Deposited:	24 Aug 2021 07:20
Last Modified:	24 Aug 2021 07:20
URI:	http://repository.its.ac.id/id/eprint/89118

Actions (login required)

View Item