Detecção de falhas em um sistema reator de tanque com agitação contínua utilizando inteligência artificial

The increasing automation of industrial processes stimulates the search for stable and reliable systems. reliable systems. However, the occurrence of failures in industrial devices is a reality, which can cause significant damage to safety, the environment and system performance. This proposes the application of the Multilayer Perceptron Neural Network as a method for detecting and diagnosing faults in a in a Continuous Stirred Tank Reactor (CSTR) system, a system that is widely used in industry. The aim is to develop a fault estimation system to act as a Fault Detection and Isolation Diagnosis (FDI) module, capable of providing information to a Fault Tolerant Control (FTC) system. This system aims to keep the CSTR operating safely and with good performance metrics, even under fault conditions. A research contributes to the evolution of methods for detecting and diagnosing faults in controlled dynamic controlled systems and highlights the essential role of Artificial Intelligence in guaranteeing reliable and safe operations in industrial process control.