Integrating Data Simulation, Machine Learning, and Mathematical Modeling for Predicting Critical Operational Parameters and Evaluating Overall Equipment Effectiveness in Food Production

Abstract

In response to various challenges faced by the food Industry, such as equipment underutilization, high-energy consumption, maintenance inefficiencies, product quality variation, waste production, and inefficient data analytics, which in aggregate represent crucial impact on productivity and operational cost. This study formulated an AI-driven optimization solution for the assessment and optimization of overall equipment effectiveness (OEE) values of machines in the food Industry for enhanced operational efficiency, productivity and profitability. The study investigated real-time sensor and machine setting data, collected from the production environment to simulate the operational conditions of the production line. A Random Forest Repressors were employed to predict main operational parameters of the production system such as downtime, defect rate, and cycle time. After being trained on a portion of the simulated data, these predictions were then utilized to compute individual components of OEE: Availability, Performance, and Quality rates and ultimately determining the overall effectiveness of the production line. The models were tested using unseen samples. A mean square error method was applied to evaluate the accuracy of the model’s predictions. The results showed that the true and expected overall equipment effectiveness (OEE) values were similar, ranging from 0.59% to 0.84%, which is relatively low compared to world-class target 85%. This result demonstrates that with the application of data-driven method to enhance the three metrics of OEE can lead to improved system operational efficiency. Additionally, the strong correlation between the actual and predicted scores signifies that the model is capable of accurately evaluating equipment effectiveness by taking multiple factors into account. The results of this article revealed that the downtime, defect rate and cycle time model performed perfectly well, while the validation results showed that the implemented AI model can effectively be explored in Industrial applications. Furthermore, the results of this study will empower maintenance engineers with the tools and insights necessary to enhance operational excellence. By enabling early identification and elimination of potential events, fostering predictive maintenance, supporting continuous improvement and enhancing collaboration within the food and other manufacturing domain.