Study On Method Of Human Reliability Analysis And Evaluation Of Coalmine Operation

According to the statistics of coalmine accidents in China in the past thirty years,the ratio of accidents caused by human errors in coalmine accidents is as high as 96.5%.Therefore,researchers try to find human factors hidden in coalmines accidents and coalmine system,identify unsafe behaviors of coalmine operating process,and explore root causes of human failure.lt is helpful for people to jump out of the perspective of coalmining accidents only from the perspective of technology and personnel skills,while taking human factors into account to avoid coalmine accidents.In order to provide theoretical support for establishing intrinsically safe mine,the new way of avoiding coal mine accidents is sought from the measures of reducing human error.Human error and human reliability is a pair of opposites,and quantitative assessment of human reliability or human error probability can lead to the root causes of human error.Though human reliability analysis has attracts more concerns from many countries and industry security researchers,human factors research has become a hot topic in the field of security research,but correlation studies mainly concentrated in the nuclear industry,aerospace and other industries with intensive technology,complex man-machine interface.The research is still in initial state,while related model is far from mature.For the coalmine system with high operation complexity and severe working environment,the research on analysis and evaluation of human reliability human reliability seems less.In view of this,this thesis studies the subject for human reliability analysis and evaluation of coalmine operations on basis of ideas and methods of existing research,combined with safety engineering,systems engineering,safety psychology,ergonomics,etc.,and some exploratory results are obtained.The main research work of this thesis are as follows:(1)Based on the comparative analysis of six major HRA methods,we combine advantages of THERP and CREAM method to establish a new analysis model named THERP+CREAM model,to make predictions of human event failure probability in coalmine system.The model applies THERP method to decompose and analyze the event,and establish human event tree.The failure probability of each sub-task of the event tree is predicted by the CREAM method,and the total failure probability of the event is finally obtained.The combination of the two methods has absorbed the advantages and improved the accuracy of overall evaluation.The formula method of CREAM environmental impact index beta is optimized to reflect the characteristics of cognitive model.The formula of failure probability is improved,and the calculation of the final failure probability reflects the influence of the cognitive pattern on human behavior.We establish the table of CPC and performance reliability,weight factor in coalmine system.Retrospective analysis cognitive failure model,the antecedent failure and retrospective table of consequent-antecedent failure in CREAM is improved,and the usage method is described in detail.(2)An indice system is constructed for human reliability evaluation of coalmine workers,which contains 9 second-level indicators,39 third-level indicators.Application of fuzzy comprehensive evaluation combined with hierarchy analysis method for human reliability evaluation of coalmine workers,can accurately acquaintance the behavior of workers,and timely detect psychological problems of workers.For specific accident prone personnel,taking promptly behavior and psychological intervention,retraining,reorientation measures,can avoid the unsafe behavior of human to the maximum extent,so as to reduce the accident occurrence.(3)For the first time,the concept and calculation formula of the observation reliability are proposed.By obtaining external performance data of human reliability of coalmine workers,the observation human reliability calculation formula is applied to solve the problem of human reliability as a latent variable,which can not be directly measured.Then put forward the four period theory of human reliability varies of workers.For four period respectively,fuzzy comprehensive evaluation and neural network methods are used to predict the human reliability.Based on BP neural network,a model is established by reference to nine vectors of human reliability evaluation indice system,which overcome the problem of using the traditional evaluation method for incomplete data samples,that is difficult to evaluate;The application of improved method of BP neural network Levenberg-Marquardt for human reliability evaluation of coalmine workers in later performance fluctuation period enhance the prediction accuracy and stability.PCA method is introduces into data reduction for the input vector set of RBF neural network using.According to contribution rate of the principal component to reconstruct the original sample space,it is effective to reduce the dimension of the input vector set,and eliminate the input correlation between each dimension and retain the main information of the original vector set.The method greatly reduces the complexity of original problem and improve the evaluation speed and the generalization ability of neural network.Based on RBF neural network,we establish an evaluation model,which is suitable for human reliability evaluation of coalmine workers in early performance fluctuation period to obtain an initial value of human reliability.RBF algorithm can overcome the problem of traditional evaluation methods encounter,that is difficult to evaluate samples lack of data,because of its fast convergence speed and high prediction precision.(4)The application of bayesian network modeling breaks through the inherent assumption in fault tree analysis,which is easy to consider multiple states and interdependency among variables.It is better to express the uncertainty relation than logic gates.The coalmine process failure can be divided into inspection,conmmunication,decision-making,operation as four behavior misfunction,and these behavior misfunction can be subdivided into different failure modes.We peopose a method which determine the CFP0 based on expert scoring method and combine it with formula method to calculate the CFP.It is clearer in failure mechanism compared with fixed CFP0 method,and solve the subjectivity problem of using the expert scoring method with broad scoring standard to determine the CFP.We carry out sensitivity analysis to key nodes of bayesian network by changing the specific factors of data,then recalculate the probability of human error event caused to quantify chain influence.By raising the human reliability of worker in key positions,it can be found that the frequency error rate shows a significant reduction trend with the increase of human reliability.The trencd can reflect the ability of personnel with high reliablility degree to shoot trouble after trigger event occurs can reduce the possibility of accident to a certain extent.The way can be used to qualify the effect of risk management,dangerous source control and position optimization.