Bayesian neural networks to reduce accidents while tower crane operations

Abstract

The subject of the article is methods of preventing and reducing the risk of an accident during the operation of a tower crane. The research focuses on the method when the Bayesian approach is integrated with an artificial neural network to determine probability distribution function of tower crane accident. The goal of the study isto provide the rational reasons why using a Bayesian neural network could reduce the level of accident risk while tower crane operating. To achieve the goal of the research were performed: review and analysis of the methods used in modern conditions to assess the level of accident risk when operating a tower crane; identified the reasons why these methods need improvements; proposed and provided justifications why Bayesian neural network could be a powerful method in preventing tower crane accident. During review and analysis were used: Bayes theorem, theory of probability and mathematical statistics, theory of artificial neural networks. The summary from the results of conducted analysis of modern methods which are used to estimate accident risk level for tower crane included that some methods are based on expert judgements only which are not possible to verify so the risk’s estimates obtained from those methods could be biased; other methods are good in estimation of the probability of happening of tower crane accident, however could not be used to prevent the accident to happen. In order to eliminated disadvantages of the used methods it is proposed to use Bayesian neural network to determine probability distribution function and use it to determine threshold value for risk factors which could provoke an accident’s event happens during tower crane operation.