RISK RAPID ASSESSMENT AT ENGINEERING ENTERPRISES THROUGH NEURAL NETWORKS

Abstract

Some issues on the risk management regarding the operation safety at the engineering enterprises are considered. The expert survey technique on the risk assessment for the mechanical engineering enterprises, and the training set are developed. The neuronet operation should be adjusted to satisfy the expert council qualification. Normally distributed differences in the individual assessments within the expert group with 3σ-interval up to 10÷12 points, or up to 100÷120 % are characteristic of the given council. The developed in the neurosimulators neuromodels for the risk analysis at the mechanical engineering enterprises have the associated with the council member’s absolute and relative error specifications. At that, the regression analysis of the training set yields essentially worse results as opposed to the neuromodel. For the construction of graphics, such software packages as STATISTICA Neural Networks and NeuroSolutions were used.