Perbandingan Model Predictive Maintenance Dengan Pendekatan Machine Learning Menggunakan Data Operasional Pabrik Semen (Studi Kasus PT. XYZ)

Wahid, Essa Abubakar (2022) Perbandingan Model Predictive Maintenance Dengan Pendekatan Machine Learning Menggunakan Data Operasional Pabrik Semen (Studi Kasus PT. XYZ). Masters thesis, Institut Teknologi Sepuluh Nopember.

Text 6032201089-Master_Thesis.pdf - Accepted Version Restricted to Repository staff only until 1 October 2024. Download (4MB) | Request a copy

Abstract

Kehandalan alat diperoleh melalui maintenance strategy yang baik. Secara umum maintenance strategy dikategorikan menjadi tiga, yaitu Corrective Maintenance (CM), Preventive Maintenance (PM), dan Predictive Maintenance (PdM). PdM adalah strategi pemeliharaan dimana aktivitas perawatan dilakukan dengan mengevaluasi dan menganalisa kondisi aktual dari alat, sehingga dapat diprediksi waktu yang optimal untuk dilakukan aktivitas perawatan sebelum kegagalan alat terjadi. Aktivitas PdM di PT XYZ dilakukan dengan memanfaatkan data real-time operasi pabrik yang terintegrasi pada Technical Information System (TIS) untuk dianalisa secara sederhana. Data parameter operasi seperti data feed rate, kecepatan revolution per minute (rpm), temperatur, tekanan udara, dan beban arus dari beberapa alat diunduh dari TIS. Data mentah kemudian dilakukan pemrosesan awal agar menghasilkan data yang siap diproses dan dimodelkan ke dalam algoritma Machine Learning seperti Random Forest, Gradient Boosting, dan XGBoosting. Selanjutnya model dievaluasi untuk diperoleh model algoritma yang paling akurat. Hasil evaluasi model Random Forest memberikan hasil paling baik di antara pemodelan lainnya dengan nilai R2 sebesar 83.1%, MSE 47.4, RMSE 6.89, MAE 3.55, dan MAPE 3.55%. Parameter operasi yang paling berpengaruh adalah actual feed rate fine coal, outlet gas temperature CT,actual rate coal weigh feeder, dan shut off flap by pass raw mill gas duct.
=====================================================================================================================================
Good equipment reliability is achieved through a good maintenance strategy. In general, the maintenance strategy is categorized into three, namely Corrective Maintenance (CM), Preventive Maintenance (PM), and Predictive Maintenance (PdM). PdM is a maintenance strategy where maintenance activities are carried out by evaluating and analyzing the actual condition of the equipment, so that the optimal time for maintenance activities can be predicted before equipment failure occurs. Predictive Maintenance activities at PT XYZ are carried out by utilizing real-time plant operation data integrated into the Technical Information System (TIS) to be analyzed in conventional way. Operational parameter data such as data feed rate, speed of revolution per minute (rpm), temperature, air pressure, and load current from several tools are downloaded from TIS. The raw data is being pre-processeded to produce data that is ready to be partitioned and modeled into Random Forest, Gradient Boosting, and XGBoosting algorithms, then tested and evaluated to obtain the most accurate algorithm model. The results of the model evaluation show that the Random Forest model gave the best results among other models with R2 83.1%, MSE 47.4, RMSE 6.89, MAE 3.55, and MAPE 3.55% The most influential operating parameters are actual feed rate fine coal, outlet gas temperature CT, actual rate coal weigh feeder, dan shut off flap by pass raw mill gas duct.

Item Type:	Thesis (Masters)
Additional Information:	RTMT 620.004 6 Wah p-1 2022
Uncontrolled Keywords:	Kehandalan, TIS, Machine Learning, Predictive Maintenance
Subjects:	Q Science > Q Science (General) > Q325.5 Machine learning.
Divisions:	Interdisciplinary School of Management and Technology (SIMT) > 61101-Master of Technology Management (MMT)
Depositing User:	- Davi Wah
Date Deposited:	01 Dec 2023 06:51
Last Modified:	01 Dec 2023 06:51
URI:	http://repository.its.ac.id/id/eprint/105238

Actions (login required)

View Item