基于改进贝叶斯神经网络的航空公司安全敏感性分析

针对当下航空公司安全质量管理体系(Quality Management System,SQMS)中风险识别与可靠性改进的问题,提出了基于区间数学改进的贝叶斯神经网络的灵敏度分析方法。利用区间数学理论分析贝叶斯神经网络中各指标与整体安全质量状况的扰动关系,实现指标灵敏度分析。通过东方航空公司的实例分析,发现在对指标进行人工干预时组合指标干预效果较好,且安全管理体系实施后指标的灵敏度有明显向好的方向变化的趋势。     

Eliciting expert knowledge in conservation science

Expert knowledge is used widely in the science and practice of conservation because of the complexity of problems, relative lack of data, and the imminent nature of many conservation decisions. Expert knowledge is substantive information on a particular topic that is not widely known by others. An expert is someone who holds this knowledge and who is often deferred to in its interpretation. We refer to predictions by experts of what may happen in a particular context as expert judgments. In general, an expert-elicitation approach consists of five steps: deciding how information will be used, determining what to elicit, designing the elicitation process, performing the elicitation, and translating the elicited information into quantitative statements that can be used in a model or directly to make decisions. This last step is known as encoding. Some of the considerations in eliciting expert knowledge include determining how to work with multiple experts and how to combine multiple judgments, minimizing bias in the elicited information, and verifying the accuracy of expert information. We highlight structured elicitation techniques that, if adopted, will improve the accuracy and information content of expert judgment and ensure uncertainty is captured accurately. We suggest four aspects of an expert elicitation exercise be examined to determine its comprehensiveness and effectiveness: study design and context, elicitation design, elicitation method, and elicitation output. Just as the reliability of empirical data depends on the rigor with which it was acquired so too does that of expert knowledge.     

基于HCNAM-BN的化工过程爆炸事故致因链分析

为了从源头上预防化工过程爆炸事故,依据风险耦合理论,探讨了各风险因子非线性耦合演化为爆炸事故的机理,构建了层次耦合网络分析模型（HCNAM）;从多因素风险耦合角度分析了国内外44起典型化工过程爆炸事故,统计了各风险因子之间的耦合概率并进行了耦合致因重要度分级;采用耦合概率与二态分布相结合的条件概率分布,将层次耦合网络分析模型转化为贝叶斯网络,并对氯乙烯单体槽爆炸性混合气体爆炸事故进行了应用研究。结果表明:91种双因子耦合风险状态中,47种呈现弱耦合致因特性;7种因子双耦合形成风险的概率较大;基于HCNAM-BN模型分析事故,可有效辨识事故最可能致因因素,获取各事故致因链的发生概率并确定事故网络关键节点。     

Physiochemical properties influencing biomass abundance and primary production in Lake Hoare, Antarctica

The perennially ice-covered, closed basin lakes in the McMurdo Dry Valleys respond rapidly to environmental changes, especially climate. For the past 15 years, the McMurdo Dry Valleys Long-Term Ecological Research (MCM-LTER) program has monitored the physical, chemical and biological properties of the lakes in Taylor Valley. In order to better assess the physiochemical controls on the biological process within one of these lakes (Lake Hoare), we have used vertical profile data to estimate depth-dependent correlations between various lake properties. Our analyses reveal the following results. Primary production rates (PPR) are strongly correlated to light (PAR) at 12-15 m and to soluble reactive phosphorus (SRP) at 8-22 m. Chlorophyll-a (CHL) is also positively correlated to PAR at 14 m and greater depths, and SRP from 15 m and greater. This preliminary statistical analysis supports previous observations that both PAR and SRP play significant roles in driving plant growth in Lake Hoare. The lack of a strong relationship between bacterial production (BP) and dissolved organic carbon (DOC) is an intriguing result of the analysis.     

Estimating influence of stocking regimes on livestock grazing distributions

Livestock often concentrate grazing in particular regions of landscapes while partly or wholly avoiding other regions. Dispersing livestock from the heavily grazed regions is a central challenge in grazing land management. Position data gathered from GPS-collared livestock hold potential for increasing knowledge of factors driving livestock aggregation patterns, but advances in gathering the data have outpaced advancements in analyzing and learning from it. We fit a hierarchical seemingly unrelated regression (SUR) model to explore how season of stocking and the location where cattle entered a pasture influenced grazing distributions. Stocking alternated between summer on one side of the pasture one year and fall on another side of the pasture the next year for 18 years. Waypoints were recorded on cattle for 50 d each year. We focused our analysis on the pasture's 10 most heavily grazed 4-ha units, because these units were the most prone to negative grazing impacts. Though grazing of the study units was always disproportionately heavy, it was much heavier with the summer than fall stocking regime: Bayesian confidence intervals indicate summer grazing of study units was approximately double the average fall grazing value. This is our core result, and it illustrates the strong effect stocking season or date or both can have on grazing distributions. We fit three additional models to explore the relative importance of stocking season versus location. According to this analysis, stocking season played a role, but stocking location was the main driver. Ostensibly minor factors (e.g. stocking location) can greatly influence livestock distributions.     

船舶溢油应急反应人的可靠性分析研究

为减小船舶溢油事故应急反应人员失误概率,提升应急处置效果,在对应急人员可靠性主要影响因素识别的基础上,利用模糊贝叶斯网络(BN)建立船舶溢油应急人的可靠性分析(HRA)模型,将应急反应流程抽象为可进行概率推算的BN,实现单个应急任务和全过程人的可靠性的量化评估,并依据该模型对一起船舶溢油事故进行实例研究。结果表明,首先模糊集值专家评估较好地解决应急人员可靠性情境依赖性强,难以量化评估的问题;其次,BN卓越的概率推理性能实现由单个应急任务向全过程人的可靠性的推算;最后得到应急反应人的可靠性主要取决于关键应急任务的完成情况和相关情境因素的影响。     

基于在线检测信息的储罐底板腐蚀状态智能评价方法

为了解决利用声发射在线检测技术对储罐底板腐蚀状态进行评价时,主要依赖检测人员经验的问题,使该项技术能更好地推广和应用,利用储罐底板在线检测的声发射信息和外观检查信息,并根据相关标准及专家经验,确定与储罐底板腐蚀状态相关的表征因素。采用遗传算法(GA)改进贝叶斯网络(BN)搜索方法,建立基于GA的BN智能评价方法。针对声发射在线检测信息和外观检查信息,分别建立基于标准用声发射因素、基于声发射因素和综合考虑声发射因素和外观检查因素的基于在线检测信息的储罐底板腐蚀状态评价模型。通过对测试样本的评价,对比声发射检测专家评价结果,其中基于在线检测信息的储罐底板腐蚀状态评价模型的准确率为96%,该模型能够对储罐底板腐蚀状态进行可靠的智能评价。     

Inverse calculation approaches for source determination in hazardous chemical releases

Source determination is vital in decision making and emergency planning involving hazardous chemical releases. This work was concentrated on inverse calculation approaches for source determination as well as current trends and future perspectives. In this paper, these different approaches are reviewed by dividing them into two categories: probability modeling methods and optimization modeling methods. The traits of these approaches are comparatively analyzed. Then it is shown how these approaches behave when applied to practical cases, and their feasibility, applicability, stability, and limitation in determining the location and strength are presented. It is argued that when experimenting with potential terrorist attacks involving hazardous chemical releases, observation points should be around the main line of the downwind direction when the source is known; while the uniform distribution of observation points is an efficient solution for unknown incidents. Probability modeling methods are demonstrated to be insufficient during emergency responses due to their lacking of enough prior information of unknown parameters, while optimization modeling methods are efficient and become a new trend in source determination. Findings reflect an urgent need for the development of high-accuracy detectors and further research of data transmission techniques in order to ensure the validity of these approaches.     

危化品事故情景推演网络构建及关键节点分析*

为有效应对危化品事故发生与演变时存在的动态性及不确定性问题,基于事故演变过程中的关键情景状态及对应孕灾环境、应急活动和处置目标等要素,引入动态贝叶斯网络方法,构建危化品事故情景推演网络,并结合复杂网络知识,对孕灾环境和应急活动2类事故影响因素节点进行敏感性和重要度分析。结果表明:情景推演网络计算的情景节点概率符合事故发生的实际情况,能够对危化品事故在不同孕灾环境和应急救援下的演变路径进行推演,并且分析得到燃烧物、消防力量、火场环境和建筑密度等关键影响因素节点,研究结果可为事故处理提供应急辅助决策支持。