Imtiez Fliss

Imtiez fliss, Moncef Tagina

Fault detection methods have their pros and cons. Thus, it is possible that some methods can complement each other and offer consequently better diagnostic systems. The integration of various characteristics is a way to develop "hybrid" systems to overcome the limitations of individual strategies of each method. In this paper a novel detection module combining the use of adaptive threshold and fuzzy logic reasoning inspired by the Evsukoff's approach is proposed in order to reduce the rate of false alarms, guarantee more robustness to disturbances and assist the operator in making decisions. The proposed approach can be used in case of multiple faults detection. This approach is applied to a benchmark in diagnosis domain: the three-tank system. The results of the proposed detection module are then presented through a gradual palette of colors in the graphical interface of the system.

Imtiez Fliss, Moncef Tagina

This work proposes to put up a tool for diagnosing multi faults based on model using techniques of detection and localization inspired from the community of artificial intelligence and that of automatic. The diagnostic procedure to be integrated into the supervisory system must therefore be provided with explanatory features. Techniques based on causal reasoning are a pertinent approach for this purpose. Bond graph modeling is used to describe the cause effect relationship between process variables. Experimental results are presented and discussed in order to compare performance of causal graph technique and classic methods inspired from artificial intelligence (DX) and control theory (FDI).

Imtiez Fliss, Moncef Tagina

In this paper an on-line multiple faults detection approach is first of all proposed. For efficiency, an optimal design of membership functions is required. Thus, the proposed approach is improved using Particle Swarm Optimization (PSO) technique. The inputs of the proposed approaches are residuals representing the numerical evaluation of Analytical Redundancy Relations. These residuals are generated due to the use of bond graph modeling. The results of the fuzzy detection modules are displayed as a colored causal graph. A comparison between the results obtained by using PSO and those given by the use of Genetic Algorithms (GA) is finally made. The experiments focus on a simulation of the three-tank hydraulic system, a benchmark in the diagnosis domain.