Integration of Seasonal Autoregressive Integrated Moving Average and Bayesian Methods to Predict Production Throughput Under Random Variables

• Main Author: Amir, Azizi
• Format: Article
• Language: English
• Published: Faculty Mechanical Engineering, UMP 2014
• Subjects: TS Manufactures
• Online Access: http://umpir.ump.edu.my/id/eprint/8238/1/Integration_of_Seasonal_Autoregressive_Integrated_Moving_Average_and_Bayesian_Methods_to_Predict_Production_Throughput_Under_Random_Variables.pdf; http://umpir.ump.edu.my/id/eprint/8238/; http://dx.doi.org/10.15282/jmes.7.2014.23.0121

Analysing and modelling efforts on production throughput are getting more complex due to random variables in today’s dynamic production systems. The objective of this study is to take multiple random variables of production into account when aiming for production throughput with higher accuracy of prediction. In the dynamic manufacturing environment, production lines have to cope with changes in set-up time, machinery breakdown, lead time of manufacturing, demand, and scrap. This study applied a Bayesian method to tackle the problem. Later, the prediction of production throughput under random variables is improved by the Seasonal Autoregressive Integrated Moving Average (SARIMA) method. The integrated Bayesian-SARIMA model consists of multiple random parameters with multiple random variables. A statistical index, R-squared, is used to measure the performance of the integrated model. A real case study on tile and ceramic production is considered. The Bayesian model is validated with respect to the convergence and efficiency of its outputs. The results of the analyses indicate that the Bayesian-SARIMA method produces a higher R-squared value, at 98.8%, compared with previous studies on Bayesian methods where the value was 90.68% and the ARIMA method where it was 97.38%. Consequently a robust approach in terms of the degree of prediction accuracy is proposed. This integrated method may be applied for the estimation of other production performance factors like lead time and cycle time in different types of dynamic manufacturing environment.