FASTa fault detection and identification software tool

Résumé

The aim of this work is to improve the reliability and safety of complex critical control systems by contributing to the systematic application of fault diagnosis. In order to ease the utilization of fault detection and isolation (FDI) tools in the industry, a systematic approach is required to allow the process engineers to analyze a system from this perspective. In this way, it should be possible to analyze this system to find if it provides the required fault diagnosis and redundancy according to the process criticality. In addition, it should be possible to evaluate what-if scenarios by slightly modifying the process (f.i. adding sensors or changing their placement) and evaluating the impact in terms of the fault diagnosis and redundancy possibilities. Hence, this work proposes an approach to analyze a process from the FDI perspective and for this purpose provides the tool FAST which covers from the analysis and design phase until the final FDI supervisor implementation in a real process. To synthesize the process information, a very simple format has been defined based on XML. This format provides the needed information to systematically perform the Structural Analysis of that process. Any process can be analyzed, the only restriction is that the models of the process components need to be available in the FAST tool. The processes are described in FAST in terms of process variables, components and relations and the tool performs the structural analysis of the process obtaining: (i) the structural matrix, (ii) the perfect matching, (iii) the analytical redundancy relations (if any) and (iv) the fault signature matrix. To aid in the analysis process, FAST can operate stand alone in simulation mode allowing the process engineer to evaluate the faults, its detectability and implement changes in the process components and topology to improve the diagnosis and redundancy capabilities. On the other hand, FAST can operate on-line connected to the process plant through an OPC interface. The OPC interface enables the possibility to connect to almost any process which features a SCADA system for supervisory control. When running in on-line mode, the process is monitored by a software agent known as the Supervisor Agent. FAST has also the capability of implementing distributed FDI using its multi-agent architecture. The tool is able to partition complex industrial processes into subsystems, identify which process variables need to be shared by each subsystem and instantiate a Supervision Agent for each of the partitioned subsystems. The Supervision Agents once instantiated will start diagnosing their local components and handle the requests to provide the variable values which FAST has identified as shared with other agents to support the distributed FDI process.