A method for determining the emergency risk level based on the integration of a cascaded artificial neural network and a vague data preprocessing block

The problem of determining the level of emergency risk based on combining machine learning methods and fuzzy data is considered. Modern systems for determining the level of risk and analyzing emergency situations require an integrated approach that will combine artificial intelligence technologies and expert knowledge. The proposed method includes several interrelated blocks of actions that ensure high accuracy of risk assessment. The division into constant and variable factors that significantly affect the risk of an accident. The block of fuzzy data preprocessing is necessary for the formation of several data sets (the opinion of the majority and the opinion of a minority of experts), with the help of which it is possible to more fully take into account the opinion of all experts. This avoids ignoring unlikely but critical risks. The weighted average expert opinion calculation block is necessary to calculate the weighted average risk level for constant and variable factors. The final risk is calculated as a weighted sum of the estimates of both groups. The coefficients of the weights (for example, 0.4 for the majority, 0.2 for the minority) are adapted to the context of the situation occurring at the facility. The object will be considered a farm with a closed-circuit water supply system. For example, in conditions of uncertainty, the weight of the minority opinion increases or the weight of the majority decreases. Advantages of the method: the integration of neural networks and expert assessments ensures adaptability to different scenarios and the ability to take into account values that can only be expressed qualitatively when determining the level of risk. Taking into account the opinions of all experts (alternative opinions) helps to increase the accuracy of the assessment of the situation and reduces the risk of “blind spots”. Cascading data processing speeds up real-time decision-making and helps avoid serious consequences of accidents. This method can be used not only on installation of a closed water supply (USV) farms, but also for other areas of industry, transport and energy, where it is necessary to use a combination of expert systems with AI systems and this is critically important for accident prevention.