Forecasting of production and scrap amounts using artificial neural networks

Abstract

The increase in consumer needs and the scarcity of production resources cause the concept of ‘productivity’ to be essential for companies. Reducing costs is an essential factor for increasing competitiveness, and therefore, businesses are taking action to reduce scrap costs and increase efficiency. Since the increase in scrap will reduce productivity, it may cause production delays and thus customer dissatisfaction. In this study, the slitting line of one of the essential Japanese supplier companies operating in the automotive sector in Turkey is discussed. The proposed model aims to predict the amount of production and scrap that may occur to increase productivity in the slitting line by using an artificial neural network (ANN) and increasing the efficiency of the slitting line with the measures to be taken. In this context, different ANN designs were made for production and scrap. During the execution of the ANN models, the production and scrap amounts were forecasted at 99 and 85%, respectively. While measuring the successful performance of the ANN models, root mean squared error, mean absolute percent error and R 2 indicators were used; the forecasted values produced by the ANNs that were successful in terms of performance indicators were compared with the actual values, and the reliability of the study was increased.