A new hybrid approach based on probability distribution and an improved machine learning for multivariate risk assessment

Abstract

A highly complex dynamic non-linear reactor is the blast furnace iron manufacturing system. It has possible dangers, including carbon monoxide, wide variety of chemical reactions, fire, high pressure and explosion, noise, split and fall, hot metal sparks, hit etc. To ensure a secure working, organizations must take the required measures to manage the risks and their effects. The approach for risk assessment discussed in this research attempts to increase blast furnace safety performance and reduce workers injuries. This approach uses probability distribution and an improved machine learning techniques such as radial basis function artificial neural networks (RBANN). The novelty here is to calculate a multivariate risk using a proposed method, namely exponential smoothing combined with radial basis function artificial neural networks (ES-RBANN). To identify their limits, the results of a research comparing conventional and novel techniques are confirmed using real data collected from the steel production operations ArcelorMittal-Annaba, Algeria.