煤矿作业人因可靠性分析与评价方法研究

发布时间：2018-01-17 02:03

本文关键词：煤矿作业人因可靠性分析与评价方法研究　出处：《大连理工大学》2017年博士论文　论文类型：学位论文

【摘要】：通过对我国近三十年来煤矿事故类型的统计可知,煤矿事故中由于人因失误造成的事故比率高达96.5%。因此,研究煤矿事故及煤矿系统中的人的因素,识别煤矿作业过程中的不安全行为,探究引发人因失误的根原因,有助于人们跳出仅从技术和人员技能的角度采取避免煤矿事故的思维,而兼顾人的因素;并从减少人因失误的措施入手寻求避免煤矿事故的新思路,为建立本质安全型矿井提供理论支持。人因失误与人的可靠性是一对对立统一体,研究人的可靠性,对人因可靠性或人因失误概率进行定量评估,可以找出产生人因失误的根本原因。尽管人因可靠性分析已得到各国家、各行业安全研究人员的密切关注,人的因素研究也成为安全研究领域的热点课题,但是各国的研究主要集中在核工业、航天等技术密集、人机界面复杂的行业,并且研究亦处于初始状态,分析和评价模型不尽成熟。对于操作复杂性高、工作环境更为恶劣的煤矿系统,人因可靠性的分析和评价研究较少。鉴于此,本文在已有研究的基础上,结合安全工程学、系统工程学、安全心理学、人机工程学等学科的思想和方法,针对煤矿作业人因可靠性分析与评价问题进行了研究,得到了一些探索性的成果。论文的主要研究工作如下:(1)建立新的HRA分析模型——THERP+CREAM模型,并改进CREAM方法。在对六种主要HRA分析方法进行比较分析的基础上,结合THERP及CREAM方法的优点,建立THERP+CREAM模型对煤矿系统的人因事件失效概率进行预测。该模型应用THERP方法对事件分解、分析,建立事件树;应用CREAM方法对事件树的各子任务的失效概率进行预测,最终得到事件的总体失效概率。两种方法的结合,吸收了各自的优点,提高了总体评价的准确性。对CREAM方法中环境影响指数β的计算公式进行了改进,使之体现了认知模式的特点;改进了失效概率计算公式,使得最终的失效概率的计算体现了认识模式对人的行为的影响;建立了煤矿系统中CPC与绩效可靠性、权重因子的关系表;对CREAM方法的追溯分析认知失效模式类型、失效前因以及"后果-前因"追溯表进行了改进,并对其使用方法进行了详细说明。(2)构建针对煤矿作业人员人因可靠性评价的指标体系。该指标体系共包含9项二级指标,39项三级指标。应用模糊综合评价结合层次分析法对煤矿作业人员人因可靠性进行评价,可更准确地了解员工的行为规律,及时发现员工的心理问题,对于特定的事故倾向人员,及时采取行为和心理干预、再培训、转岗等措施,最大限度地避免人的不安全行为,从而减少人因事故的发生。(3)建立煤矿作业人员人因可靠性评价的指标体系的神经网络评价模型并对人因可靠性进行预测分析。首次提出观测可靠度概念及计算公式。通过获取煤矿作业人员人因可靠性的外在表现数据,应用观测可靠度计算公式,解决了人因可靠性作为一个潜变量,无法直接度量的问题;提出了作业人员可靠度变化的四时期理论,针对四个时期分别应用模糊综合评价及神经网络方法进行了人因可靠性预测。参照人因可靠性九大评价指标体系建立的基于BP神经网络的评价模型,克服了应用传统评价方法对数据不全、样本难以评价的困难;采用改进的BP神经网络方法Levenberg-Marquardt对处于后绩效波动期的煤矿作业人员进行人因可靠性评价,提高了预测精度和稳定性;运用PCA方法对RBF神经网络的输入向量集进行约简,根据贡献率取其主成分重构原样本空间,有效地降低了输入向量集的维度,消除了输入各维度之间的相关性,保留了原向量集的主要信息,大幅度降低了问题复杂度并提高了评价速度和神经网络的泛化能力;建立了基于RBF神经网络的评价模型,该模型对处于前绩效波动期的煤矿作业人员进行人因可靠性评价,并赋予可靠性初值,RBF算法具有收敛速度快、预测精度高的特点,克服了传统评价方法对数据不齐备样本难以评价的困难。(4)应用贝叶斯网络方法建模,解决多工种共同作业人因事故概率预测问题,并对特定人因事件的关键岗位进行敏感度分析。将煤矿作业失效分为检查、交流、决断、操作四种行为功能失效,并对各类行为功能失效进行了失效模式的细分;基于专家打分法确定CFP0,并结合公式法计算CFP的方法,与固定CFP0方法相比,失效机理更加清晰,也解决了专家打分法确定CFP主观性过强,打分标准过于宽泛的问题;应用贝叶斯网络方法建模突破了故障树分析中固有的假设,能够考虑变量的多态性和变量间的相关性,比逻辑门更好地表达变量间的不确定性关系。对贝叶斯网络的关键节点进行敏感性分析:通过改变特定影响因素的数据,将其对人误事件造成的影响定量化;通过调高关键岗位作业人员的可靠性,可以发现频度失误率随着岗位可靠度的提高而呈显著降低的趋势,能够体现高可靠度作业人员对于触发事件发生后的事故修正能力可以在一定程度上降低事故发生的可能性,能够实现危险源管控和岗位优化的效果。
[Abstract]:Based on the statistical data of the type of coal mine accidents in China in the past thirty years, coal mine accidents due to human error caused by the accident rate as high as 96.5%. therefore, factors of coal mine accidents in coal mine system and people's unsafe behavior, identification of mine in the process of operation, for inquiry triggered root cause failure, help in people jump out only from the technical and skill point taken to avoid coal mine accidents and thinking, both human factors; and from reducing human error measures seek to avoid the new idea of coal mine accidents, for the establishment of intrinsically safe mine. To provide theoretical support for human reliability failure and people are a pair of opposites one, the study of human reliability, the reliability or quantitative assessment of error probability, can find out the root cause for produce errors. Although the human reliability analysis has been in various countries, various industries. Pay close attention to the researchers, the study of human factor has become a hot topic in the field of security research, but the research mainly concentrated in the nuclear industry, aerospace and other technology intensive, man-machine interface, complex industry, and is also in the initial state, analysis and evaluation model is not mature. For the operation of the high complexity of work environment for the coal mine system bad, human reliability analysis and evaluation research. In view of this, on the basis of existing research, combined with the safety engineering, system engineering, safety psychology, thoughts and methods of ergonomics and other subjects, according to the mine operation for reliability analysis and evaluation are studied, get some exploration the results. The main research work of the thesis are as follows: (1) the establishment of a new HRA model -- THERP+CREAM model, and the improved CREAM method. In the analysis of the six major HRA Based on the comparative analysis method, combining the advantages of THERP and CREAM method, the paper establishes the THERP+CREAM model for coal mine system for event failure probability prediction. This model using the THERP decomposition method, analysis of the event, the establishment of event tree; failure probability using CREAM method of event tree sub task prediction, finally obtained the event total failure probability. The combination of the two methods, absorbing their advantages, improve the accuracy of the overall evaluation of the CREAM method. The environmental influence index formula of beta was improved, which embodies the characteristics of cognitive model; improved the calculation formula of failure probability, the failure probability of the final show the understanding of the impact on human behavior; establish CPC performance and reliability of coal mine system, the relationship between the weight factor table; back to the CREAM method analysis of cognitive failure mode The type of failure, and antecedent consequent antecedent retrospective table "has been improved, and its use in detail. (2) to construct the index system for coal mine workers for reliability evaluation. The index system includes 9 indicators in level two and 39 indicators in level three. Evaluation of the operating personnel coal mine reliability analysis with fuzzy comprehensive evaluation combined with hierarchy analysis method, can more accurately understand the behavior of employees, timely find the psychological problems of employees, for the staff to take the specific accident tendency, behavior and psychological intervention, retraining, reorientation and other measures, the maximum to avoid unsafe behaviors, so as to reduce the because of the accident. (3) neural network established the evaluation index system for reliability evaluation model of the coal mine workers and forecast analysis on human reliability analysis. First proposed by the concept of degree of observability and gauge Formula. Through the acquisition of coal mine workers due to external performance data reliability, calculation formula of application of observation reliability, solve the human reliability as a latent variable, cannot directly measure the problem; put forward the reliability change operator theory of the four period, for four respectively during the period of application of fuzzy comprehensive evaluation and nerve the network method of human reliability prediction. The evaluation based on BP neural network model reference to establish evaluation index system for reliability of nine, which overcame the problem of using the traditional evaluation method for incomplete data is difficult to sample evaluation difficulties; using the improved method of BP neural network Levenberg-Marquardt in the coal mine workers after the performance of phase fluctuation human reliability assessment, improve the prediction accuracy and stability; the input vector of RBF neural network set reduction using PCA method, according to contribution The rate of the principal component reconstruction of the original sample space, effectively reduce the input vector dimension, eliminating the input correlation between the various dimensions, retain the main information of the original vector set, greatly reduce the complexity of the problem and improve the generalization of neural network can evaluate the speed and force; establish the evaluation RBF neural network the model based on the model of human reliability evaluation in coal mine workers before the performance fluctuation period, and gives the reliability of the initial value, the RBF algorithm has fast convergence speed and forecasting precision, overcome the traditional evaluation method for incomplete data sample is difficult to evaluate the difficulties. (4) the application of Bayesian network modeling method. To solve many types of common operation prediction problem because of the accident probability, and the specific sensitivity analysis for key positions. The coal mine operation failure event is divided into examination, communication, decision, operation four The failure behavior and the behavior of all kinds of function, function failure is a failure mode of subdivision; determine the CFP0 based on the method of expert scoring method and CFP was calculated with formula method, compared with the fixed CFP0 method, failure mechanism more clearly, but also solve the expert scoring method to determine the CFP subjective scoring standard is too strong, too broad the application of Bayesian network modeling; fault tree analysis methods to break the assumptions inherent in, can consider the correlation between variables and polymorphism between variables and variables than the expression of the uncertainty relations between logic gates better. Sensitivity analysis of key nodes of Bayesian networks: by changing the specific influencing factors of quantitative data of its impact on human error events; the reliability of the adjustable high key personnel, can find the frequency error rate with post reliability increased and significantly decrease It can reflect high reliability workers' ability of correcting accidents after triggering events, which can reduce the possibility of accidents to a certain extent, and achieve the effect of risk source control and post optimization.

【学位授予单位】：大连理工大学
【学位级别】：博士
【学位授予年份】：2017
【分类号】：TD82;TD77

【参考文献】