BATOUT ,NaoualBENDIB, Riad2024-04-282024-04-282024-02-15http://dspace.univ-skikda.dz:4000/handle/123456789/1437Petrochemical installations are complex processes with high connectivity and continuously interaction between their systems and devices, which requires the control of several parameters for optimal operation. This type of process deals with hazardous substances at elevated temperatures and pressure and always faces difficulties in process control, hence failing to target operations that generate several types of risks such as explosion and fire. These safety incidents lead to serious consequences affecting people, environment and property. To provide a high level of security and safety to these installations, different risk management strategies have featured in literature, to identify and reduce hazardous situations. However, even with the vast variety in these strategies and the combination of different techniques to deal with some limitations and cover the maximum drawbacks, other difficulties are still considered such as uncertainties in both the input and output of the analysis and the classification problem. In recent years several tools have been developed based on artificial intelligence to deal with these difficulties such as fuzzy logic that relies on membership function principle and artificial neural network that emulate the biological ones. The objective of the thesis is to define generic methods of risk analysis based on fuzzy logic and neural network that can be appropriated for any industrial system and evaluate the different risks very well under the condition of uncertainty.  The first: an integrated frame implemented based on HAZOP and fuzzy evaluation of risk matrix to evaluate the safety integrity level (SIL) of an industrial fired heater safety.  The second: a novel methodology based on LOPA method and FUZZY LOGIC to increase its performance in terms of analysis and risk reduction. This approach is implemented to a real system namely naphtha-A- stabilizer after identifying risks inherent in this system by applying the HAZOP method (Hazards and Operability Study).  The third: an approach based on Artificial Neural Networks (ANN) is developed to schedule the SIL values of the safety integrity functions (SIF) of an industrial-fired heaterenThesis