Forecasting for Steel Production using Artificial Neural Networks and Feasibility Analysis of Plant Regeneration Acid Development in PT. XYZ

Abstract

PT. XYZ is a steel production industry company. The Acid Regeneration Plant (ARP) facility has now shown a decrease in plant availability. This study aims to determine steel demand forecasting and also conduct a feasibility study for the development and replacement of ARP technology. Artificial Neural Networks, Linear Regression and Decomposition are used for forecasting experiments. The choice of forecasting method is taken by looking at the smallest MAPE value of each method. Artificial Neural Networks have the smallest MAPE value of 2.58. Then the Artificial Neural Network is used in forecasting requests for 2019 to 2020. The chosen network architecture is 24-12-1 with the traingdx training function. The results obtained are an increase in demand for steel products from 2019 to 2020. The increase in demand is 1, 329, 398 tons, an increase of 4.59% from the previous 2 years. Then a feasibility study is conducted to assess the construction of new ARP facilities. The feasibility study covers technology selection and investment appraisal. After making a comparison, the technology used is Spray Roaster. Based on the results of the investment appraisal using the NPV, IRR PBP, the new ARP development investment is decent and can meet future steel demand in the period until 2020.