FORECASTING MODEL DEVELOPMENT FOR BREAKDOWN PREDICTOR FACTOR OF INJECTION MOLDING MACHINE

• Main Author: FAHRI, AKHMAD
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/46970
• Institution: Institut Teknologi Bandung

Packaging Service Division (PSD) Cibitung is a part of PT X and Cibitung Plant which produces preforms, semi-finished polyethilene terephtalate (PET) based carbonated soft drink bottles, using injection molding machines. PSD Cibitung suffers from a high amount of unscheduled downtime which in turn causes the machine availability to be low at 66,16% on average. The high amount of unscheduled downtime is due to many cases of machine breakdowns which occur at the production lines. This shows the ineffectiveness of maintenance activities and policy in PSD Cibitung to prevent machine breakdowns. Hence, PSD Cibitung decided to develop condition-based maintenance (CBM) policy. Therefore, this research aims to discover the appropriate predictor variables for machine breakdowns and develop forecasting models for said variables using IRIS PdM methodology in order to assist the CBM policy development. This research has compared three classifier algorithms as the means to discover the appropriate predictor variables for machine breakdowns with recall as performane metric to minimize false negative. Decision tree is selected as the best classifier algorithm with 75% recall. The decision tree classifier has shown that part and oil temperature are the appropriate predictor variables for machine breakdowns. Oil temperature is a metric variable which changes over time so it is decided to design a forecasting model for oil temperature. This research has compared some forecasting models and selected autoregressive integrated moving average (ARIMA) models as the forecasting models for oil temperature. A forecasting system prototype has been developed to assist in the decision making on maintenance treatments based on the variable condition forecast.