Improving operational performance using prediction algorithms: A case study at an air filter plant - Challenge Site

Abstract

The study aims to employ prediction algorithms to analyse, evaluate and improve operational performance quality indicators at the air filter factory/challenge site, which were represented by (defect rate, on-time delivery rate, operational efficiency, overall equipment effectiveness, operating cycle time, specification compliance rate, and resource consumption rate). The importance of the study lies in identifying the basic deviations in performance quality indicators and then predicting the likelihood of success in the coming years. The study followed an analytical methodology using algorithms to analyse (logistic regression, decision tree, Random Forest, VSM) that predict the future by analysing current data, using Python, which works on a set of libraries specialising in algorithm analysis. The Pandas library was used for data processing and cleaning, and the Scikit-learn library to apply supervised learning algorithms, in addition to the Matplotlib and Seaborn libraries to visually represent the data and analyse the results. The fundamental problem is manifested in increased unplanned downtime, higher defect rates, resource depletion, and reduced equipment reliability resulting from the operational characteristics of intermittent production patterns, where frequent changeovers and start-up and shutdown cycles impose mechanical and technical stresses that lead to repeated unplanned downtime. This imbalance is structurally reflected in the quality of operational performance. The results indicate that statistical analysis and intelligent analysis unanimously agree that 2015 was the best year in terms of performance quality, while the remaining years were classified as poor performers.