Transferability of accident prediction models

Accident prediction models, the vast majority of which are negative binomial regression models, are of considerable importance to highway agencies since they can be used to conduct many traffic safety studies. However, not every agency possesses sufficient accident statistics that enable it to develop reliable models of its own. This problem gives rise to interest in the transferability of accident prediction models in time and space. It would save time, effort, and money if accident prediction models developed for one region in one period of time could be applied in different time periods and regions to produce reliable safety studies.This paper presents methods for recalibrating negative binomial accident models before transferring them for use in different time periods and regions of space. The paper emphasizes that the recalibration of the shape parameter of a transferred model using local data is absolutely necessary. It explains that it is also desirable to recalibrate the constant term of the transferred model in order to allow the model to better suit local conditions. A moment method is presented for recalibrating the shape parameter of a transferred model when its constant term is not recalibrated. However, a maximum likelihood method is presented for recalibrating both the shape parameter and the constant term of the transferred model and is shown to be superior to the recalibration methods existing in the traffic safety literature.     

Solving the seveso II prediction problem of substance generation during loss of control by Al techniques.

The European Directive 96/82/EC forces industries to know exactly what substances may be generated during loss of control of an industrial chemical process. The study must be as accurate as possible, as the design of the safety measures that will be ready in case a real loss of control occurs depends on these results.Chemical reactions under out of control conditions, follow defined but difficult to know rules, specially at high temperatures showing high molecular complexity. It seems that the application of artificial intelligence techniques will help to predict which substances are produced given an initial chemical scenario. In this paper the k-nearest algorithm and its application to this problem will be studied.A database has been developed to manage reactive information independently from the algorithm or technique of AI to be applied. The most important information of the database are relations between substances present on the initial reactive scenario and the ones detected after reaction under out of control conditions occurred. Chemical substances contained in the database have been analysed by decomposing them into Benson’s Groups, which is thought to be the most adequate level of detail to preserve chemical information while establishing relations between non-identical substances.The Shepard’s modification of k-nearest neighbor algorithm has been used for the Boolean prediction of formation of a certain substance given a reactive scenario. Results show an accuracy above 95%.     

Safety line method for the prediction of deep coal-seam gas pressure and its application in coal mines

Gas pressure is an important index for evaluating the outburst risk and determining the gas content in coal seams. It is recommended to predict coal-seam gas pressure of the workface at deep levels before extending mining activities to deeper levels. According to the prediction results, measurements are taken for gas outburst prevention and control and for workload estimation. At present, regression methods are always used to process the numerous gas pressure data for prediction. Because there are many factors that influence the gas pressure which could lead to a deviation from actual values, the measured data do not possess basic conditions for regression methods; this can cause unexpected dangers if the methods are adopted.Based on a statistical analysis of actual measured results of coal-seam gas pressure in a same geological section in certain coal mine, two symbol measured points are selected to make a line for prediction, i.e. safety line, and the other measured points should be below the line except the abnormal points due to the confined water. It has been successfully applied in numerous coal mines in China. Particularly, this method is analyzed in this paper for the case of the No. 8₂ coal seam in the Taoyuan coal mine in Huaibei coalfield, China. By comparatively analyzing the relationship between gas pressure and depth from surface using regression methods, it is found that the safety line method could lead to a better prediction for deep coal-seam gas pressure, and therefore promote early warning ability and mining safety.     

A research model--forecasting incident rates from optimized safety program intervention strategies

INTRODUCTION/PROBLEM: Property damage incidents, workplace injuries, and safety programs designed to prevent them, are expensive aspects of doing business in contemporary industry. The National Safety Council (2002) estimated that workplace injuries cost $146.6 billion per year. Because companies are resource limited, optimizing intervention strategies to decrease incidents with less costly programs can contribute to improved productivity. METHOD: Systematic data collection methods were employed and the forecasting ability of a time-lag relationship between interventions and incident rates was studied using various statistical methods (an intervention is not expected to have an immediate nor infinitely lasting effect on the incident rate). RESULTS/SUMMARY: As a follow up to the initial work, researchers developed two models designed to forecast incident rates. One is based on past incident rate performance and the other on the configuration and level of effort applied to the safety and health program. Researchers compared actual incident performance to the prediction capability of each model over 18 months in the forestry operations at an electricity distribution company and found the models to allow accurate prediction of incident rates. IMPACT ON INDUSTRY: These models potentially have powerful implications as a business-planning tool for human resource allocation and for designing an optimized safety and health intervention program to minimize incidents. Depending on the mathematical relationship, one can determine what interventions, where and how much to apply them, and when to increase or reduce human resource input as determined by the forecasted performance.     

A beetle antennae search improved BP neural network model for predicting multi-factor-based gas explosion pressures

A gas explosion in an underground structure may cause serious damage to the human body and ground buildings and may result in huge economic losses. The pressure of the gas explosion is an important parameter in determining its severity and designating an emergency plan. However, existing empirical and computational fluid dynamics (CFD) methods for pressure prediction are either inaccurate or inefficient when considering multiple influencing factors and their interrelationships. Therefore, for a more efficient and reliable prediction, the present study developed a multifactorial prediction model based on a beetle antennae search (BAS) algorithm improved back propagation (BP) neural network. A total of 317 sets of data which considered factors of geometry, gas, obstacle, vent, and ignition were collected from previous studies. The results showed that the established model can predict pressures accurately by low RMSE (43.4542 and 50.7176) and MAPE (3.9666% and 4.9605%) values and high R² (0.7696 and 0.7388) values for training and testing datasets, respectively. Meanwhile, the BAS algorithm was applied to improve both the calculation efficiency and the accuracy of the proposed model by enabling a more intelligent hyperparameter tuning method. Furthermore, the permutation importance of input variables was investigated, and the length (L) and the ratio of length and diameter (L/D) of geometry were found to be the most critical factors that affect the explosion pressure level.     

Fused CFD-interpolation model for real-time prediction of hazardous gas dispersion in emergency rescue

Accidental releases of hazardous gases in chemical industries can pose great threats to public security. The computational fluid dynamics (CFD) model is commonly applied to predict gas dispersion in complex structured areas. It can provide good accuracy but it is too time-consuming to be used in emergency response. To reduce computation time while keep acceptable accuracy, this paper proposes several fused CFD-interpolation models which combine CFD model with different interpolation methods. Spline, linear and nearest interpolation methods are used. A CFD simulations database is created ahead of time which can be quickly recalled for emergency usage and unknown situations can be predicted instantly by interpolation methods instead of time-consuming CFD model. Fused models were applied to a case study involving a hypothetical propane release with varying conditions and validated against CFD model. The validation shows that prediction accuracy of these fusion models is acceptable. Among these models, CFD-Spline interpolation model performs best. It is faster than CFD model by a factor of 75 and is potentially a good method to be applied to real-time prediction.     

BP神经网络法预测唐山市需水量

需水量预测研究已成为当前水资源规划与管理研究中的重要课题之一.本文设定不同的神经网络运行次数,根据预测结果进行误差分析,BP神经网络在运行5 000次时,具有高度的可信度和可行性.应用5 000次运行次数的BP神经网络模型对唐山市规划水平年的需水量进行预测.最后引入人均综合用水量概念,结果表明,预测结果在理论上和实际上都具有可行性.     

Mathematical methods for application of experimental adiabatic data – An update and extension

The paper represents some results of comparative analysis of the methods used for processing and interpreting data of adiabatic calorimetry as well as applying it to practical situations. Specifically two approaches are compared – approximate method based on evaluation of simplified kinetics and a more comprehensive, simulation-based method that utilizes the evaluation of more detailed kinetic models.The analysis is focused on two important types of data processing – correction of experimental results on thermal inertia (phi-factor correction) and estimation of adiabatic time to maximum rate (TMR).The most widely cited method for phi-factor correction is considered and its improvement is proposed to enable more precise prediction of the adiabatic time scale. A procedure for phi-factor correction of pressure response is also proposed. The limitations of this enhanced Fisher's method are discussed by comparison with simulation-based method. All the illustrative materials are based on real examples.As an example of application, the simplified method will be used to predict TMR and its limitations will be discussed.     

Evaluation of a CFD porous model for calculating ventilation in explosion hazard assessments

In the past, gas explosion assessment relied on worst case scenarios. A more realistic approach is to look at the probability of explosions and their likely severity. The most flexible way of investigating many different scenarios is to estimate a ventilation flow, feed this into a flammable volume calculation and then calculate the explosion severity. The procedure allows many parameters to be varied efficiently. A Computational Fluid Dynamics porous model is evaluated for modelling the ventilation flow through congested regions, including a new method that has been developed to derive the resistance. Comparison with velocity measurements from a large scale model of an offshore module showed that overall the CFD model performs very well, especially considering that the homogenous porosity block does not model any of the internal obstructions and therefore would not predict any local flow effects. This gives confidence that the overall flow pattern is sufficiently close to the local flow patterns, to be used in explosion assessments. The porous approximation in CFX is found to underpredict the turbulence intensity in the obstacle array compared to the explosion model EXSIM. Improving the turbulence prediction in the porous model would be valuable, so a relatively simple method of increasing the turbulence in porous regions is proposed. The CFD model will provide the non-uniform natural ventilation flowfields of complex regions for future explosion assessments at a hierarchy of levels.     

Improved far-field blast predictions from fast deflagrations,DDTs and detonations of vapour clouds using FLACS CFD

The CFD tool FLACS was developed from 1982 with a primary goal to predict gas explosion loads inside oil platform modules. The prediction of far-field blast loads was of secondary importance as any scenario creating a substantial far-field blast would already have collapsed the module where it originated. For the same reason the potential for a deflagration-to-detonation-transition (DDT) was not initially of interest. Over the past decade use of FLACS has been more widespread, and the tool is now frequently used to predict explosions on onshore facilities and FPSOs/FLNGs, where far-field blast loads and evaluation of DDT potential may be of significant interest. Previous work by Hansen et al. (2010) has highlighted a weakness in FLACS when predicting the far-field blast from strong gas explosions and, when using FLACS according to guidelines, far-field blast pressures will often be significantly underpredicted. For scenarios involving DDT this effect will be particularly strong. The current study will present a way to obtain more accurate far-field blast predictions by modified parameter settings in FLACS for strong deflagrations. Using modified settings, it is also possible, with good precision, to predict flame speeds, pressures and far-field blast from DDT-scenarios and directly initiated gas detonations, physics which are beyond the accepted capabilities of FLACS. Selected full-scale experiments from the DNV GL test site at Spadeadam will be used to compare with the simulations. Convincing evidence for DDT in large scale natural gas experiments (91% methane) was found in simulations of one of these tests.     

The role of the controller in the accelerating industry of air traffic management

This paper considers the role of the controller in future Air Traffic Management (ATM), an industry which is undergoing considerable and rapid change at this time. In particular, the paper focuses on the area of allocation of function, i.e. the determination of what the (human) controller should do, what the hardware and software (machine) should do, and what tasks they should share, and who (or what) is in control, in this increasingly complex system. The premise of the paper is that traditional criteria for allocation of function, so-called Fitts List approaches, are no longer clear-cut to apply, if they ever were. Technology is reaching the point where many traditionally human functions and roles can be supported or even completely autonomously carried out by automated systems. The question is quite simple — what functions, roles, and even responsibilities should be automated? This question is becoming an imperative in currently accelerating technologies such as ATM. As an example, aviation saw a huge insurgence of automation into the cockpit, with four generations of so-called glass cockpit designs, culminating in aircraft which are completely fly-by-wire, and where some aircraft have envelope protection such that the pilot's control actions are monitored and may be overridden by the machine. However, this transition from largely manual flight to glass-cockpit control was not without problems and automation-assisted accidents, and the air traffic industry would do well to avoid such problems. ATM is an area that has been relatively non-automated for the past 30 years. But as traffic levels continue to rise rapidly, there is general agreement that ATM systems must adopt some level(s) of automation support in order to maintain safety and efficiency of air traffic operations. However, ATM development is currently also highly technology-driven, perhaps with most emphasis on what technology can do for us, rather than what we need it to do. From a human factors perspective, the question of what we need from technology and automation is the critical one. As technology continues to accelerate, it is probable that it will be able to deliver whatever functions we wish it to. It is therefore appropriate to consider what the role of the human should be. If such considerations are not made now, then accelerating technology and traffic levels, and the need for ATM systems that can cope, will deliver systems which, if they fail, will lead to the types of negative experiences and fatal accidents that the aviation world has already experienced. Quite simply, if the role of the human is not considered now, it will be too late to consider it. Furthermore, if this occurs, then future as yet unforseen automation-assisted accidents will inevitably be attributed to human error. This paper considers some of the factors and issues surrounding the difficult area of automation and allocation of function, and gives an example of one method, based on error analysis, which has been used to try to answer some of the difficult automation questions currently facing ATM and human factors.     

Implementing and evaluating novel safety training methods for construction sector workers: Results of a randomized controlled trial

Introduction: The construction industry is regarded as one of the most unsafe occupational fields worldwide. Despite general agreement that safety training is an important factor in preventing accidents in the construction sector, more studies are needed to identify effective training methods. To address the current research gap, this study evaluated the impact of novel, participatory safety training methods on construction workers’ safety competencies. Specifically, we assessed the efficacy of an immersive virtual reality (VR)-based safety training program and a participatory human factors safety training program (HFST) in construction industry workplaces. Method: In 2019, 119 construction sector workers from eight workplaces participated in a randomized controlled trial conducted in Finland. All the study participants were assessed using questionnaires at baseline, immediately after the intervention and at one-month follow-up. We applied generalized linear mixed modeling for statistical analysis. Results: Compared to lecture-based safety training, VR-based safety training showed a stronger impact on safety motivation, self-efficacy and safety-related outcome expectancies. In addition, the construction sector workers who participated in the VR-based safety training showed a greater increase in self-reported safety performance at one-month follow-up. Contrary to our study hypotheses, we found no significant differences between the study outcomes in terms of study participants in the HFST training condition and the comparison condition without HFST training. Conclusion: Our study indicates that VR technology as a safety training tool has potential to increase safety competencies and foster motivational change in terms of the safety performance of construction sector workers. In the future, the efficacy of participatory human factors safety training should be studied further using a version that targets both managerial and employee levels and is implemented in a longer format. Practical implications: Safety training in virtual reality provides a promising alternative to passive learning methods. Its motivating effect complements other safety training activities.     

电解铝生产环境负荷分析和预测模型研究

利用生命周期评价的分析方法,对铝电解生产过程中资源消耗、能源消耗和污染物排放进行了分析,采用等效环境指数计算了铝电解生产过程中的环境负荷,并分析了各因素对环境负荷的影响,其中氟化盐的投入量对环境负荷影响较大.运用神经网络对铝电解生产过程的环境负荷进行预测,在负荷预测过程中,首先对样本数据进行归一化处理,然后采用BP算法对神经网络进行训练.最后用训练好的网络进行预测,将预测结果与实际数据进行比较,证明具有较好的预测效果.     

基于BP网络理论的岩爆预测方法

选取影响岩爆的一些主要因素,如地应力大小、岩石抗压和抗拉强度、岩石弹性能量指数,采用人工神经网络理论,根据国内外一些岩石地下工程实例构造样本集,建立了一种新的岩爆预测模型.此模型可以直接应用于岩石地下工程,对岩爆的发生与否及烈度大小进行预测.实例表明,预测结果与实际情况符合得很好,说明了此模型的有效性.     

Simple method for predicting pressure behavior during gas explosions in confined spaces considering flame instabilities

It is indispensable to predict the pressure behavior caused by gas explosions for the safety management against accidental gas explosions. In this study, a simple method for predicting the pressure behavior during gas deflagrations in confined spaces was examined. Previously the pressure behavior was calculated analytically assuming laminar flame propagation. However, the results of this method often provide underestimation compared with experimental data. It was known the underestimation intensifies as the scale of explosion spaces becomes larger. On the large scale gas deflagration, flame instability (especially hydrodynamic instability) might be more effective and wrinkles appeared on the flame front. Then, the flame surface area was increased and the propagating flame was gradually accelerated. The ordinary prediction methods led to the underestimation because the propagating flame was assumed to be laminar. In this study, we considered the effect of flame wrinkles caused by flame instabilities. By regarding the flame front as a fractal structure, the flame surface area could be modified. Because a flame surface starts to be wrinkled on a certain flame radius, proper determination of the critical flame radius provided accurate prediction of pressure behavior on a large scale deflagration. In addition, correction of the K_G value in a large vessel was discussed.     

基于自记忆模型的煤与瓦斯突出电磁辐射预测研究

利用实验测定的电磁辐射信号时间序列,用双向差分原理反导出一个非线性常微分方程;以其为微分动力核,运用动力系统数据机理自记忆模式构造自记忆方程并求出自记忆系数;利用该方程预测未来电磁辐射信号的变化,并与现场测定对比分析,用误差分析和距平分析法验证该模型正确性和预测准确率。实例表明:该自记忆模型预测与实测结果是一致的,相对误差均在6.7852%左右,距平符合率为90%;自记忆方法能有效应用于煤与瓦斯突出电磁辐射动态预测中;该模型与电磁辐射预测方法的有机结合能有效地提高预测准确性,从而为煤与瓦斯突出电磁辐射预测技术提供了一种新的研究途径。     

人工神经网络在煤与瓦斯突出预测中的应用

由于煤与瓦斯突出发生机理的复杂性,传统预测方法的应用受到很大的限制,而人工神经网络理论以其高度非线性映射的特性为解决这一问题提供了新的途径。以突出预测指标为基础,利用多层反向传播神经网络(BP网络)模型实现对突出危险性的预测。实例分析表明,模型精度很高,可用于工作面煤与瓦斯突出危险性的预测。     

SVM application in hazard assessment: Self-heating for sulfurized rust

In order to assess the oxidation self-heating hazard of sulfurized rust, for particular ambient conditions in crude oil tanks, the support vector machine (SVM) technique is applied to predict the maximum temperature (T_max) of oxidation self-heating process. Five governing parameters are selected, i.e. the water content, mass of sulfurized rust, operating temperature, air flow rate and oxygen concentration in the respiratory/safety valve. The efficiency and validity of the SVM predictions are investigated in the case of two sets of data: more than 85 experiments performed in academic lab (China) and almost 17 additional results collected from existing literature. Two main steps are also discussed: the training process (on selected subsets of data) and prediction process (for the remaining subsets of data). It can be concluded that for both datasets the maximum temperature (T_max) values calculated by SVM technique were in good accordance with the experimental results, with relative errors smaller than 15% except for a few cases.The SVM technique seems therefore to be relevant and very helpful for complex implicit processes such as chemical reactions, as it is the case of the oxidation of sulfurized rust in oil tanks. Furthermore, such predictive methods can be continuously be improved through additional experiments feedback (larger databases) and can then be of crucial help for monitoring and early warning of hazardous reactions.     

GRNN模型在煤与瓦斯突出及瓦斯含量预测中的应用

煤与瓦斯突出的作用机理非常复杂,是诸多因素如地应力、煤层瓦斯、煤体物理力学性质等共同作用的结果。在分析广义回归神经网络(GRNN)的基本原理和算法的基础上,建立煤与瓦斯突出等级以及基于构造复杂程度定量评价的瓦斯含量GRNN模型。然后用收集到的工程实例样本训练和检验该模型。结果表明,GRNN模型具有很好的预测能力和泛化能力,能较好揭示瓦斯含量和诸影响因素间的关系,可用于煤与瓦斯突出判别以及瓦斯含量预测。同时可以看出,光滑因子的合理选取对于提高GRNN模型的预测精度非常重要,因此,在以后的实际应用中需要不断尝试,找出最合理的光滑因子。     

GIS技术在瓦斯动力灾害预测中的应用

GIS系统是一种重要的空间信息系统,可以采集、存贮、管理、描述和分析与空间和地理分布相关的数据,适时地提供多种空间和地貌的信息.本文结合平顶山矿区的生产实际,将GIS技术引入煤与瓦斯突出区域的预测中,建立相应的信息管理系统,提高了信息处理水平.