Observer-Based Fault Detection System for a Class of Industrial Process Implemented with MQTT Protocol

Fault detection systems supporting industrial safety and reliability are essential for increasing the performance and cost-effectiveness of the process. In the case of automatic control, a traditional way is to improve the reliability of the individual components of the system, such as sensors and actuators. Although currently data-driven approaches are very popular, the data availability is not completely guaranteed. On the other hand, the mathematical model of some technical processes is friendly to develop. In this case, observer-based approaches are suitable to generate residual signals stating the existence of faults. Under certain conditions, fault diagnosis is also possible. Additionally, with the growing of Industry 4.0, MQTT is currently one of the most used protocols for implementing IIoT applications. This paper tackles the problem of a very common industrial process related to the level control of cascade tanks. This class of process is highly available in the chemical, food, pharmaceutical industries, water distribution systems, and usually its mathematical model is easy to configure. In our study, the model was configured according to the characteristic of the physical prototype available in our laboratory, to simulate seven faults in sensors, actuator, tanks, and valves. All the faults were modelled, and a bank of observers-based filters were designed to detect and diagnose all the seven faults through the residual signals. A real scenario of real time processing was simulated by deploying the time-discrete version of the fault diagnostic system in a software application implemented under MQTT protocol.