Study On Human Reliability Based On RBF Neural Network

As a result of modern science and technology development,the reliability of machine has been increased tremendously,but the person has become the most unreliable factor of man-machine system relatively,which is one of the elements of man-machine system. Statistical results of the contemporary accidents have sufficiently testified that the human errors have already become the main reason of accident. Meanwhile,the modern industrial system is becoming more and more hugely and complicatedly,once accident happened,the damage caused by the accident is extremely serious. Therefore it is necessary to study the reasons and the patterns of human errors and the measures of controlling for the safety of modern industry.This dissertation systemically analyzed feasibility and superiority of the PCA-RBF Neural Network in the Human Reliability Analysis, which based on Principal Component Analysis (PCA) and Human Reliability Analysis (HRA) firstly. Secondly, with the analysis of factors influencing human reliability, the basic rules that constructing the forecasting system of human operating reliability should obey and the main contents that the forecasting model should include are chosen firstly. Then take the height operating workplaces which is one of the"easy-lead-accident-place"in the construction enterprises as an example,based on quantization of the factor of human reliability,with the dimensionality reduction theory of PCA and the technology of the " visualization of thinking" to RBF neural network, the human reliability of height operators is forecasted qualitatively and the question of selecting is resolved before working. Lastly,take a crane operator as an example, the human reliability during the working is forecasted quantitatively, then controlling human errors in the course of operating has been implemented.By validating the example, the model could implement the human error prediction commendably. The traditional HRA method has difficulty about forecasting assignments decomposition and reliability evaluate of action in a complex and mobility operation. Under the similar conditions, taking the times of human errors as the index of forecasting human reliability, the dissertation can resolve this difficulty. The quantitative forecasting model of the human reliability can resolve the dynamic forecasting problem of human reliability which the traditional HRA couldn't resolve at a certain workplace. Meanwhile, the model combines the human-action-reliability with the human-preservation-reliability successfully and improves the accuracy of the forecast results.The safety system engineering, human factors engineering, reliability engineering, the RBF neural network, principal component analysis and the MATLAB are banded together organically to take to HRA predict, the model is helpful to select the operators and control the important or serious dangerous positions or monitoring positions of the major hazards ,such as chemical plant,mine,construction and so on, and then enterprise safety management level can be enhanced.