This paper investigates real-world data from a PVC manufacturing plant in Saudi Arabia to construct predictive statistical models leveraging machine learning techniques. The primary aim is to identify prevalent failures and predict their timing based on historical incidents. The study introduces the Random-Forest-Classifier algorithm to refine the dataset and enhance accuracy. Subsequently, the results are applied to simulation modeling, providing insights into proactive action and opportunistic maintenance behavior within PVC manufacturing. The motivation of the research was to reduce the sudden breakdown in the factory and provide practical recommendations to optimize maintenance practices, thereby enhancing operational efficiency. The paper concludes with a simulation model illustrating the use of opportunistic actions that support the Overall Equipment Efficiency (OEE) resulting from the predictive model's insights.