Optimizing Predictive Maintenance in Smart Factories Using Random Forests and Internet of Things Sensors

In the rapidly evolving landscape of Industry 4.0, smart factories leverage advanced technologies such as the Internet of Things (IoT) and machine learning to enhance operational efficiency. Predictive maintenance (PdM) is a critical aspect of this transformation, aiming to anticipate and mitigate equipment failures before they occur. However, existing PdM models often suffer from high false positive rates and inadequate scalability. This study addresses these issues by proposing an optimized predictive maintenance framework utilizing Random Forests and IoT sensors, tailored specifically for smart factory environments. Our methodology integrates a comprehensive IoT sensor network within the factory, collecting real-time data on various parameters such as temperature, vibration, and machine utilization. This data is preprocessed to handle noise and outliers, followed by feature extraction to identify the most significant predictors of equipment failure. We employ an optimized Random Forest algorithm, enhanced with a novel dynamic threshold adjustment technique and an adaptive learning mechanism to improve prediction accuracy and model robustness. The model’s performance is evaluated using a large dataset collected from multiple manufacturing units over a period of six months. The proposed framework significantly outperforms traditional PdM approaches. Our optimized Random Forest model achieved an accuracy of $\mathbf{9 6. 8 \%}$, a precision of $\mathbf{9 4. 5 \%}$, and a recall of $\mathbf{9 7. 2 \%}$, compared to ${88.3 \%, 85.7 \%}$, and 89.1% respectively for the baseline models. Additionally, the dynamic threshold adjustment technique reduced the false positive rate by 40%, while the adaptive learning mechanism enhanced the model’s ability to generalize.