Factories are currently confronted with multifaceted challenges created by rapid technological Many technologies have recently appeared and evolved, including Cyber-Physical Systems, the Internet of Things, Big Data, and Artificial Intelligence. Companies established various innovative and operational strategies, there is increasing competitiveness among them and increasing companies’ value. A smart factory has emerged as a new industrialization concept that exploits these new technologies to improve the performance, quality, controllability, and transparency of manufacturing processes. Artificial intelligence and Deep Learning techniques are revolutionizing several industrial and research fields like computer vision, autonomous driving, predicting failures, etc. The idea of this work is the development of a predictive model to predict the amount of raw material in a workshop in a cement factory based on the Deep Learning technique Long Short-Term Memory (LSTM). The excellent experimental results achieved on the LSTM model showed the merits of this implementation in the production performance, ensuring predictive maintenance, and avoid wasting energy.

Industrial systems are difficult to control and supervise efficiently because of the complexity of the production process. The aim is to automatically control in real-time as an alternative for operators as possible and highlight the importance of machine learning in the field of industry. Integrating SVM into the industrial supervision system in the cement factory (SCIMAT) permits the classification of different measurements coming from sensors to the Programmable Logic Controller (PLC) that indicates when the process is in good functioning or bad indicating that a default has occurred. These measurements are classified after training in three classes of level (low, medium, and high) that are classified in their turn into two classes (good and bad functioning). The three classes present the inputs of the fuzzy controllers. Based on this classification, the PLC makes orders for industrial equipment. Then a regression of variation of measurements in real-time is carried out to predict the good or the bad functioning of the production line. In conclusion, the proposed approach innovates the complex supervision system to learn how to control and preserve the habitual linguistic language used by operators, react in the right way, and prevent critical situations.