Validation and data splitting in predictive regression modeling of honing surface roughness data

Model validation is critical in predicting the performance of manufacturing processes. In predictive regression, proper selection of variables helps minimize the model mismatch error, proper selection of models helps reduce the model estimation error, and proper validation of models helps minimize the model prediction error. In this paper, the literature is briefly reviewed and a rigorous procedure is proposed for evaluating the validation and data splitting methods in predictive regression modeling. Experimental data from a honing surface roughness study will be used to illustrate the methodology. In particular, the individual versus average data splitting methods as well as the fivefold versus threefold cross-validation methods are compared. This paper shows that statistical tests and prediction errors evaluation are important in subset selection and cross-validation of predictive regression models. No statistical differences were found between the fivefold and the threefold cross-validation methods, and between use of the individual and average data splitting methods in predictive regression modeling.     

A method to relate chemical accident properties and expert judgements in order to derive useful information for the development of Environment-Accident Index

The environment consists of a variety of different compartments and processes that act together in a complex system that complicate the environmental risk assessment after a chemical accident. The Environment-Accident Index (EAI) is an example of a tool based on a strategy to join the properties of a chemical with site-specific properties to facilitate this assessment and to be used in the planning process. In the development of the EAI it is necessary to make an unbiased judgement of relevant variables to include in the formula and to estimate their relative importance. The development of EAI has so far included the assimilation of chemical accidents, selection of a representative set of chemical accidents, and response values (representing effects in the environment after a chemical accident) have been developed by means of an expert panel. The developed responses were then related to the chemical and site-specific properties, through a mathematical model based on multivariate modelling (PLS), to create an improved EAI model. This resulted in EAI(new), a PLS based EAI model connected to a new classification scale. The advantages of EAI(new) compared to the old EAI (EAI(old)) is that it can be calculated without the use of tables, it can estimate the effects for all included responses and make a rough classification of chemical accidents according to the new classification scale. Finally EAI(new) is a more stable model than EAI(old), built on a valid base of accident scenarios which makes it more reliable to use for a variety of chemicals and situations as it covers a broader spectra of accident scenarios. EAI(new) can be expressed as a regression model to facilitate the calculation of the index for persons that do not have access to PLS. Future work can be; an external validation of EAI(new); to complete the formula structure; to adjust the classification scale; and to make a real life evaluation of EAI(new).     

The situation of hazardous chemical accidents in China between 2000 and 2006

From the aspects of the total quantity of accidents, regional inequality, enterprises scale and environmental pollution accidents, this study makes an analysis of hazardous chemical accidents in China for the period spanning from 2000 to 2006. The following results are obtained: firstly, there were lots of accidents and fatalities in hazardous chemical business, i.e., the number of casualty accidents fluctuated between 200 and 600/year, the number of fatality fluctuated between 220 and 1100/year. Secondly, the accident rate in developed southeast coastal areas, e.g., Guangdong, Zhejiang and Jiangsu, was far higher than that in the northwest regions, e.g., Xizang, Xinjiang, and Qinghai. Thirdly, nearly 80% of dangerous chemical accidents had occurred in small and medium-sized enterprises (SMEs). Finally, various sudden environmental pollution accidents resulted from hazardous chemicals were frequent in recent years, causing a huge damage to human and property. Then, based on the readjustment of economic structure in the last decades, the development status of Occupational Health and Safety (OHS) in SMEs and other factors, the paper explores the main causes, which offers valuable insight into measures that should be taken to reduce hazardous chemical accidents.     

Simulation of the cumulative effects of chemical spills using a spatial-temporal dynamics analysis algorithm

Accidents in urban areas involving chemical spills demands development of not only feasible emergency strategies, but also a consistent framework to protect the environment and prevent accidents. This can be possible only by a sound understanding of the environmental impact of spills and their potential long-term effects. Furthermore, the impact assessment of chemical spills can not be done disregarding the spatial-temporal pattern of previous exposures reciprocally influenced by both chemical and environmental properties. In this context, this paper presents an analysis framework to quantify the cumulative effects of chemical spills at any given point of a certain area based on a "present" history of exposure coupled with chemical and environmental properties to predict possible scenarios of future exposure and estimate in advance potential alarming levels of pollution. In the present circumstances when increasing knowledge is required for an accurate prediction of spill migration through unsaturated soil, this paper proposes an algorithm capable of incorporating models of increasing complexities to simulate the single-spill events once new advancements in the field are taken. The algorithm developed is illustrated using a simple model with homogenous and steady-state conditions to simulate the single-spill events. A hypothetical case study was constructed to illustrate the analysis steps and the benefits of the algorithm.     

危化品事故连续消耗型应急物资调度模型

应急物资调度是危化品事故应急救援工作的关键内容,危化品事故会对周边区域造成环境污染,其应急物资调度问题应该考虑多个目标。针对危化品事故应急物资调度特征,建立了连续消耗应急物资调度多目标模型。模型以缺货损失最小、调运时间最短和调运总环境风险最小为优化目标,考虑应急中心物资数量限制和救援时间限制等约束条件,采用NSGAⅡ遗传算法求解。对于求解结果,采用逼近理想解的排序方法进行选择。算例计算结果表明,多目标模型优化方案比单目标模型优化方案有较大的优势,同时说明算法合理可行。     

A goal oriented approach to software testing

There is a tendency to distribute the testing and validation effort in a software project uniformly over all system functions. However, to improve systems reliability and safety, testing effort must be focused on the functions with the highest failure consequences. This paper describes a method that computes the number of test cases given the accepted risk levels for each function. Input to the method are the total set of functions for the system, the set of possible accidents and their consequences plus the subset of accidents that can be caused by a failure in one particular function. Failure consequences and functions usage are then used to find the Potential Annual Loss Exposure (PALE) by using a simple diagram developed by AFSC. Given the PALE value, system functions can be ranked according to their risk. It is then possible to set goals for their failure probabilities and to compute the number of test cases needed for each function.     

Accidents,often the result of an ‘uncontrolled business process’—a study in the (Dutch) chemical industry

Often companies in the (petro‐) chemical industry claim that all possible countermeasures against potential accidents have been taken and therefore accidents are unforeseeable. In this paper we question this statement by analysing the pre‐warning signals (precursors) preceding a number of industrial accidents. 17 accidents that occurred in the (petro‐) chemical industry have been investigated by exploring FACTS, an accident database containing information about industrial accidents worldwide. This paper will demonstrate that the existence of precursor information could have been used to foresee and even prevent these accidents if a proper control action had been initiated. The accidents are analysed further, according to a control model, which was adapted from that of C. Argyris. It demonstrates the ineffectiveness of several elements of business process control loops and that the so‐called ‘double‐loop learning’ cycle is more important than the ‘single‐loop learning’ cycle if one considers safety improvement. Copyright © 2003 John Wiley & Sons, Ltd.     

Accidents in the construction industry in the Netherlands: An analysis of accident reports using Storybuilder

As part of an ongoing effort by the Ministry of Social Affairs and Employment of the Netherlands, a research project is being undertaken to construct a causal model for occupational risk. This model should provide quantitative insight into the causes and consequences of occupational accidents. One of the components of the model is a tool to systematically classify and analyse reports of past accidents. This tool ‘Storybuilder’ was described in earlier papers. In this paper, Storybuilder is used to analyse the causes of accidents reported in the database of the Dutch Labour Inspectorate involving people working in the construction industry. Conclusions are drawn on measures to reduce the accident probability. Some of these conclusions are contrary to common beliefs in the industry.     

Chemical Accident Simulation Tool (CAST): A System for Assessing Consequences of Accidents in Chemical Process Industry

The development and validation of a new software named chemical accident simulation tool (CAST) is presented. CAST enables development of scenarios of different types of accidental fires and explosions that can occur in chemical process industry (CPI). CAST is also capable of assessing the likely consequences of such accidents in terms of the area impacted and the types of impacts. The distinguishing features of CAST are (a) it incorporates a larger set of established models than handled by existing packages to simulate a wider variety of accidents in CPI; (b) it is developed with an integrated mapping tool to display damage zones around accident center; this makes the application useful in decision making; and (c) it calculates the results in a fast and reliable manner. Due to these attributes, CAST has increased efficiency, better understanding of the accident scenarios, and better communication of results. Validation of the software has been done with published results which shows that the codes for calculating the impacts from accidents are correct.     

Accidents in family forestry's firewood production

Firewood is commonly used around the world, but little is known about the work involved in its production and associated accidents. The objectives were to identify relationships between accidents and time exposure, workers’ age and sex, equipment used and work activities in family forestry's firewood production. Data from a postal survey in Northern Sweden were compared to a database of injuries in the same region. Most accidents occurred to 50–69 year old men, who also worked most hours. No significant differences in sex and age were found between expected and recorded accident frequencies when calculated from total work hours; however, when calculated using numbers of active persons significant differences were found for both age and sex. Frequency of accidents per unit worked time was higher for machine involving activities than for other activities. Accidents that occurred when using wedge splitter machines were responsible for most of this overrepresentation. Fingers were the most commonly injured body parts. Mean accident rate for the equipment used was 87 accidents per million work hours, and the rate was highest for wedge splitters (122 accidents per million work hours). Exposure to elevated risks due to violation of safety procedures is discussed, as well as possible preventative measures.     

Incompatibilities of chemicals

Chemical incompatibilities are potentially significant problems where hazardous chemicals are found. A number of chemical segregation systems have been developed which provide recommendations for the separation of incompatible chemicals. Three segregation systems were identified in this study: the UN Dangerous Goods system (which uses physical hazard as the main reason for segregation and has 14 categories), the US CHRIS system (which uses chemical reactivity and has 24 categories) and a third system which uses environmental risks (and has 25 categories). These systems were combined. Merging of each system was initially problematic, but became considerably easier once certain characteristics had been defined (such as flammability or water incompatibility). This gave a final merged incompatibility table containing 100 different segregation groups. This research study showed that it was possible to combine different segregation systems based on different criteria and that more comprehensive segregation systems can be developed. These can be of use in the decision-making process about where groups of chemicals may be used, and during the use of chemicals, where chemicals should not be combined. The use of more comprehensive segregation systems will also assist in developing proper measures for their control.     

Genetic algorithm for informative basis function selection from the wavelet packet decomposition with application to corrosion identification using acoustic emission

Chemical process installations are exposed to aggressive chemicals and conditions leading to corrosion. The damage from corrosion can lead to an unexpected plant shutdown and to the exposure of people and the environment to chemicals. Due to changes within and on the surface of materials subjected to corrosion, energy is released in the form of acoustic waves. This acoustic activity can be captured and used for corrosion monitoring in chemical process installations. Wavelet packet coefficients extracted from the acoustic activity have been considered to determine whether corrosion occurs, and to identify the type of corrosion process, at least for the most important corrosion processes in the chemical process industry. Feature subset selection is then applied to these wavelet coefficients to achieve a much higher accuracy in the identification of different corrosion processes than when no feature subset selection is applied to the acoustic waves. However, due to the statistical dependencies that potentially exist between the wavelet coefficients, the latter should not be selected independently from each other. Local discriminant basis selection algorithms do not take the statistical dependencies between wavelet coefficients into account. In this paper, we have used several mutual information-based approaches that take these dependencies into account and compared them to the wavelet-specific local discriminant basis selection algorithm. Furthermore, a hybrid filter-wrapper genetic algorithm, which uses a relevance-redundancy approach as a local search procedure, was designed. The highest classification accuracies are obtained with the hybrid filter-wrapper genetic algorithm, for all classifiers used in this paper. Furthermore, the proposed algorithm easily outperformed one of the most commonly used classifiers in chemometrics: partial least squares discriminant analysis (PLS-DA). A naïve Bayes classifier that uses the features selected by the hybrid filter-wrapper genetic algorithm was able to identify the absence of corrosion, uniform corrosion, pitting and stress corrosion cracking, with an accuracy of up to 87.20%.     

Differentiation between COVID‐19 and bacterial pneumonia using radiomics of chest computed tomography and clinical features

To develop and validate an effective model for distinguishing COVID‐19 from bacterial pneumonia. In the training group and internal validation group, all patients were randomly divided into a training group (n = 245) and a validation group (n = 105). The whole lung lesion on chest computed tomography (CT) was drawn as the region of interest (ROI) for each patient. Both feature selection and model construction were first performed in the training set and then further tested in the validation set with the same thresholds. Additional tests were conducted on an external multicentre cohort with 105 subjects. The diagnostic model of LightGBM showed the best performance, achieving a sensitivity of 0.941, specificity of 0.981, accuracy of 0.962 on the validation dataset. In this study, we established a differential model to distinguish between COVID‐19 and bacterial pneumonia based on chest CT radiomics and clinical indexes.     

Scenario analysis of freight vehicle accident risks in Taiwan

This study develops a quantitative risk model by utilizing Generalized Linear Interactive Model (GLIM) to analyze the major freight vehicle accidents in Taiwan. Eight scenarios are established by interacting three categorical variables of driver ages, vehicle types and road types, each of which contains two levels. The database that consists of 2043 major accidents occurring between 1994 and 1998 in Taiwan is utilized to fit and calibrate the model parameters. The empirical results indicate that accident rates of freight vehicles in Taiwan were high in the scenarios involving trucks and non-freeway systems, while; accident consequences were severe in the scenarios involving mature drivers or non-freeway systems. Empirical evidences also show that there is no significant relationship between accident rates and accident consequences. This is to stress that safety studies that describe risk merely as accident rates rather than the combination of accident rates and consequences by definition might lead to biased risk perceptions. Finally, the study recommends using number of vehicle as an alternative of traffic exposure in commercial vehicle risk analysis. The merits of this would be that it is simple and thus reliable; meanwhile, the resulted risk that is termed as fatalities per vehicle could provide clear and direct policy implications for insurance practices and safety regulations.     

A two-scale system to identify environmental risk of chemical industry clusters

Recent reform policies in China have spurred rapid industrial development. This has led to a large increase in chemical accidents, which may have catastrophic impacts on the local population and environment. As industrial facilities become more complex, it becomes more difficult to control and mitigate the risks associated with chemical accidents. In this study, we propose a two-scale system for assessing the environmental risk level of chemical industry clusters. A series of risk early warning indices for both the plant-specific level and regional clusters level are used in this system. Firstly, at the enterprise scale, a risk early warning index is constructed using inputs such as the presence of hazardous materials, the operation of critical plant equipment and the efficiency of extant management techniques. Secondly, an index for quantifying risks on regional scales depends on environmental, economic, and social conditions as well as the specific enterprises' components. As an illustration, the system is applied to a case study involving a five-plant chemical industry cluster in Jiangsu province, China. A geographical information system-based methodology is used to obtain a composite index score for each mesh of the five plants. The results prove that the proposed two-scale early warning system can efficiently identify environmental risk and help guide emergency responses at both the enterprise and cluster level.     

Characterizing hydraulic fracturing fluid <Emphasis Type="Italic">greenness</Emphasis>: application of a hazard-based index approach

Growth of the unconventional gas industry is predicted to continue to be an important component of the global energy landscape. The rapid expansion of shale and tight gas development has raised many environmental and human health concerns, particularly in regards to ground and surface water contamination. The unconventional gas industry has begun to transition toward the use of hydraulic fracturing chemicals that pose minimal environmental and human health hazards in order to mitigate the risks associated with possible chemical containment failure. Integrated chemical hazard evaluation has been facilitated by an adapted index-based approach to combine noncommensurate multiparameter chemical hazard data into a single score value. Comparative analysis of existing chemical hazard index scoring systems as well as the formulation of a novel hydraulic fracturing fluid greenness assessment system revealed several important considerations for index development and application. Index scores calculated using the investigated index systems highlighted the need for informed, optimized hazard class selection as input for score determination, the maintenance of hazard category intensity during parameter transformation, as well as representative hazard class and chemical component mathematical weightings, and robust aggregation techniques for final score calculation. Continued research should work to model the combined hazard posed by individual chemicals while considering the effect of dilution as well as incorporate additional index metrics beyond hazard intensity. Fully disclosed index systems, applied with complete knowledge of their strengths and weaknesses, provide useful monitoring and communication tools to promote environmental-best practices in the unconventional gas industry.     

Forward and backward risk assessment throughout a system life cycle using dynamic Bayesian networks: A case in a petroleum refinery

In this paper, risk modeling was conducted based on the defined risk elements of a conceptual risk framework. This model allows for the estimation of a variety of risks, including human error probability, operational risk, financial risk, technological risk, commercial risk, health risk, and social and environmental risks. Bayesian network (BN) structure learning techniques were used to determine the relationships among the model variables. By solving a bi-objective optimization problem applying the genetic algorithm (GA) with the Pareto ranking approach, the network structure was learned. Then, risk modeling was performed for a petroleum refinery focusing on HydroDeSulfurization (HDS) technology throughout its life cycle. To extend the model horizontally and make it possible to evaluate the risk trend throughout the technology life cycle, we developed a dynamic Bayesian network (DBN) with three-time slices. A two-way forward and backward approach was used to analyze the model. The model validation was performed by applying the leave-one-out cross-validation method.     

Identification of fireproofing zones in Oil&Gas facilities by a risk-based procedure

Fire is among the more dangerous accident scenarios that may affect the process and chemical industry. Beside the immediate and direct harm to workers and population, fire may also cause damages to structures, which may trigger escalation resulting in severe secondary scenarios. Fireproofing is usually applied to improve the capacity of structures to maintain their integrity during a fire. Past accidents evidenced that the available standards for fireproofing application in onshore chemical and process plants do not consider all the fire scenarios that may cause structural damage. In the present study a methodology was developed for the identification of the zones where fireproofing should be applied. The effect of both pool fires and jet fires was accounted. Simplified criteria, based on radiative heat intensity, were provided for the identification of the fire protection zones. A risk-based procedure was proposed for the selection of significant reference release scenarios to be used in the evaluation of worst credible fire consequences.     

Applying quantile regression for modeling equivalent property damage only crashes to identify accident blackspots

Hot spot identification (HSID) aims to identify potential sites—roadway segments, intersections, crosswalks, interchanges, ramps, etc.—with disproportionately high crash risk relative to similar sites. An inefficient HSID methodology might result in either identifying a safe site as high risk (false positive) or a high risk site as safe (false negative), and consequently lead to the misuse the available public funds, to poor investment decisions, and to inefficient risk management practice. Current HSID methods suffer from issues like underreporting of minor injury and property damage only (PDO) crashes, challenges of accounting for crash severity into the methodology, and selection of a proper safety performance function to model crash data that is often heavily skewed by a preponderance of zeros. Addressing these challenges, this paper proposes a combination of a PDO equivalency calculation and quantile regression technique to identify hot spots in a transportation network. In particular, issues related to underreporting and crash severity are tackled by incorporating equivalent PDO crashes, whilst the concerns related to the non-count nature of equivalent PDO crashes and the skewness of crash data are addressed by the non-parametric quantile regression technique. The proposed method identifies covariate effects on various quantiles of a population, rather than the population mean like most methods in practice, which more closely corresponds with how black spots are identified in practice. The proposed methodology is illustrated using rural road segment data from Korea and compared against the traditional EB method with negative binomial regression. Application of a quantile regression model on equivalent PDO crashes enables identification of a set of high-risk sites that reflect the true safety costs to the society, simultaneously reduces the influence of under-reported PDO and minor injury crashes, and overcomes the limitation of traditional NB model in dealing with preponderance of zeros problem or right skewed dataset.     

A combined M5P tree and hazard-based duration model for predicting urban freeway traffic accident durations

The duration of freeway traffic accidents duration is an important factor, which affects traffic congestion, environmental pollution, and secondary accidents. Among previous studies, the M5P algorithm has been shown to be an effective tool for predicting incident duration. M5P builds a tree-based model, like the traditional classification and regression tree (CART) method, but with multiple linear regression models as its leaves. The problem with M5P for accident duration prediction, however, is that whereas linear regression assumes that the conditional distribution of accident durations is normally distributed, the distribution for a “time-to-an-event” is almost certainly nonsymmetrical. A hazard-based duration model (HBDM) is a better choice for this kind of a “time-to-event” modeling scenario, and given this, HBDMs have been previously applied to analyze and predict traffic accidents duration. Previous research, however, has not yet applied HBDMs for accident duration prediction, in association with clustering or classification of the dataset to minimize data heterogeneity. The current paper proposes a novel approach for accident duration prediction, which improves on the original M5P tree algorithm through the construction of a M5P-HBDM model, in which the leaves of the M5P tree model are HBDMs instead of linear regression models. Such a model offers the advantage of minimizing data heterogeneity through dataset classification, and avoids the need for the incorrect assumption of normality for traffic accident durations. The proposed model was then tested on two freeway accident datasets. For each dataset, the first 500 records were used to train the following three models: (1) an M5P tree; (2) a HBDM; and (3) the proposed M5P-HBDM, and the remainder of data were used for testing. The results show that the proposed M5P-HBDM managed to identify more significant and meaningful variables than either M5P or HBDMs. Moreover, the M5P-HBDM had the lowest overall mean absolute percentage error (MAPE).