Application of Deep Learning Techniques for the Optimization of Industrial Processes Through the Fusion of Sensory Data

Abstract In the era of Industry 4.0, mastering the management and analysis of large volumes of data has become essential to foster technological advances and improve operations. Our work presents a methodology combining spatial and temporal data with deep learning techniques, specifically Convolutional Neural Networks, Recurrent Neural Networks, Long Short-Term Memory and Gated Recurrent Units, and Deep Neural Networks. We aim to transform anomaly detection, perform predictive maintenance, and optimize industrial processes. Our work has found that intelligent integration of multiple data sources improves accuracy and other key indicators, such as F1 score and AUC, and enriches decision-making with more profound, detailed information about the operating environment. The results have been promising. We have seen an increase in anomaly detection accuracy by up to 92%, an improvement in early detection for predictive maintenance by 150%, and an improvement in operational efficiency from 70% to 85%. These advances validate our proposal and demonstrate its practicality in various industrial environments. This work proposes a guide for integrating data fusion technologies in Industry 4.0, highlighting the practical benefits of our methodology and opening new possibilities for innovation and improving operational efficiency.