Sistema monitoramento de falhas aplicado em uma planta didática

Abstract

Industry 4.0 enabling technologies stand out as essential tools for optimizing and increasing the efficiency of companies in their production processes. Among its primary uses is to reduce impacts generated from industrial failures, whether of instruments, equipment, mechanics, or production. This work comprises the diagnosis of pneumatic faults in a SMAR educational industrial plant. Using OPC communication and Matlab software, data were collected in two scenarios, the first under standard operating conditions and the second with failure emulation. Applying the K-Best feature selector, the intelligent system adopted to classify patterns (failure and non-failure) was analyzed in different scenarios. In summary, through data intrinsic to the process represented by the didactic plant, they allowed the training of artificial neural networks capable of indicating faults with accuracy close to 100%. The sample data results suggest the possibility of identifying anomalous situations in the process. It is concluded that using technologies that make up Industry 4.0, such as artificial neural networks, is a viable alternative for preventive diagnosis, meeting the demands of a contemporary industry and contributing to reducing productive impacts.