With the advent of the artificial intelligence (AI) era, the chemical industrial park cyber-physical system (CIPCPS) has been widely regarded as an effective strategy for promoting the intelligent upgrading of traditional chemical industrial parks. However, it is accompanied by both opportunities and challenges. Traditional chemical industry risk assessment methods mainly focus on physical space, thereby overlooking the risks arising from the cyber space. To ensure the safety and security of CIPCPS, this study explores the following questions: (i) What emerging risks is CIPCPS currently facing? (ii) Regarding these risks, what are the current research advancements in assessment methods, and what limitations do they have? (iii) How to conduct an integrated risk assessment for CIPCPS? The results show that current research on CIPCPSs has not yet provided a systematic classification of various risk types and specific risk factors. Gaps are identified in the traditional risk assessment methods, including insufficient multi-source data fusion capabilities, limited algorithmic interpretability, and the failure to consider the unique consequence amplification effect of domino accidents in CIPCPSs. To address these issues, this study classifies the specific risk types and factors through historical accident case analysis. On this basis, an integrated risk evolution framework for CIPCPSs is developed, dividing the process into three phases: Disturbance-System induction, System-Accident evolution, and Accident-System feedback. Finally, four recommendations are proposed to guide the conduct of integrated risk assessment, providing a theoretical basis for future risk management of CIPCPSs.

Domino effects triggered by fire are responsible for most of the cascading accidents which occurred in the chemical and process industry. Their occurrence is highly dependent on the performance of fire protection barriers, as well as on the primary fire scenario. Active and passive safety barriers are largely used to prevent or mitigate domino effects. However, a blinkered allocation approach of safety measures based on a subjective strategy, which does not consider cascading effects possibly triggered by an accident, can lead to a deployment of a safety plan that is not effective in effectively mitigating the consequences of a domino accident. In the present study, an innovative methodology supporting and optimizing the decision-making process for fire prevention in a chemical site has been developed. The methodology provides an optimal allocation of safety measures, aimed at preventing and mitigating fire escalation under budget constraints. A Cost-Effectiveness Analysis with an optimization algorithm is used to allocate a limited budget in terms of fire safety barriers. A novel tool, named Cost Variable Decision Tree, has been developed to support fire prevention investment decisions. The methodology is able to assess the economics of safety plan implementation, considering both the costs of safety measures and the hypothetical benefits deriving from avoided accidents. An illustrative case-study was carried out, confirming that the proposed methodology can effectively support the decision-making process addressing safety barriers allocation.

Pipelines are the most widely used system for transporting liquid and gaseous energy materials, but throughout their lifespan, they are exposed to various detrimental factors, such as corrosion and deviations in process variables. In recent years, the concept of resilience has gained significant attention as a means to analyze infrastructure behavior during failure states. This study introduces a novel metric for assessing pipeline resilience based on reliability. The proposed method involves an aging study of pipelines, considering the interaction of potential failures—such as corrosion, pressure variations, temperature fluctuations, and changes in fluid velocity—and subsequently analyzes ways to restore the system to its original conditions. The method offers an assessment approach for the three phases that constitute a resilience curve: absorption, adaptation, and restoration. This approach not only identifies the system's time to failure, but also through analysis of the resilience curve, facilitates the comparison of the effects of potential preventive, mitigative, and repair actions. A case study is presented to validate the method's efficacy. The results suggest that the proposed approach could be a valuable tool in the decision-making process within the asset integrity management (AIM) framework, aiming to optimize pipeline resilience by implementing the most effective safety solutions.

Explosion-induced domino accidents in the chemical industry, such as the 2005 Buncefield and 2019 Xiangshui accidents, can lead to catastrophic losses. Recent studies commonly use probit models (simplified linear regression models) to predict the probability of accident escalation caused by equipment failure due to overpressure conditions but necessitate distinct equations for different equipment types. In order to simplify the number of models and improve their accuracy, this study introduced three machine learning models (random forest model, convolutional neural network model, and deep neural network model), addressing complex nonlinear relationships that conventional regression models may not fully capture. By model training, the DNN model has the highest accuracy (99 %), followed by CNN (94 %) and random RF (95 %). The DNN model was selected as the optimal data-driven model for equipment vulnerability assessment due to their feedforward mechanism's capability to dynamically align parameters with evolving data distributions. The approach developed can not only predict the probability of equipment damage by integrating values related to peak overpressure and equipment type but also effectively address the accuracy validation issues associated with traditional regression models. Besides, this approach can be considered open source model and more explosion data may be used in the future to further improve the model.

Ammonia is a promising fuel for decarbonizing the shipping industry, but its safety compared to traditional maritime fossil fuels remains questionable. As more projects explore ammonia-fueled vessels, safety concerns, particularly the acute toxicity and risks of ammonia release, are paramount. This research focused on the effects of accidental ammonia releases during ship-to-ship collisions, affecting ammonia tank integrity. We examined various release scenarios, considering factors like tank types and locations, breach sizes and positions, weather conditions, and dispersion patterns, using PHAST software for modeling. Results indicated that semipressurized tanks pose greater health risks on human health than fully refrigerated ones. Underwater releases are less hazardous, as a significant amount of ammonia dissolves before surfacing. Mitigation efforts, such as water curtains and containment basins, were evaluated for their effectiveness in minimizing the impact of ammonia releases. These measures significantly reduce risks to nearby populations but are less effective for crew safety onboard. This underscores the challenge of ensuring onboard safety in ammonia-fueled vessels, highlighting the need for innovative and effective safety design.

Emergency response is a procedural safety barrier of paramount importance for the mitigation of fire scenarios and the prevention of escalation. However, in Natech scenarios, emergency response may be affected by the natural event impacting the site. Indeed, when contrasting Natech accidents, emergency responders have to face both the natural event and the cascading technological scenario. Despite the criticality of the issue, limited attention was devoted to date to the analysis of emergency response in cascading sequences triggered by natural events. The present study provides a novel and technically sound methodology to assess the performance of emergency response and the required intervention time in Natech scenarios. An expert survey combined with a Bayesian Network model was used to assess the performance of the emergency response. The routing and setup phases were identified as those mostly affected by natural events. Monte Carlo simulations were used to obtain baseline data and specific probability distributions for the time required to carry out the emergency response considering the factors that may hinder the response during natural events. In Natech accidents, the time for effective mitigation resulted higher of at least a factor 2 with respect to that expected in the case of conventional accidents. The methodology developed may be used to support the improvement of the emergency management of Natech scenarios, allowing for a detailed definition of site-specific emergency response plans. Moreover, the results may be used to provide a more accurate assessment of the fire-driven escalation probability in Natech events.

Deliberate attacks (security attacks) pose a significant threat to offshore Oil&Gas critical infrastructures as they have the potential of triggering major event scenarios with severe consequences on people, property, and the surrounding environment. The standards API RP 70 and API RP 70I address security issues in the offshore Oil&Gas sector, providing a semi-quantitative approach to evaluate the actual level of security risk. However, as the credibility of security attacks grows, security risk assessments should be approached in a more systematic and quantitative way to measure vulnerabilities and determine the level of protection available in the site. In this context, the present study introduces a systematic quantitative procedure using Bayesian Network (BN) to calculate the probability of success of physical attacks and the role of preventive and mitigative response strategies. The procedure is applied to a case study allowing to show its potential for improving security in the offshore Oil&Gas industry.

Fire is one of the main accident scenarios occurring in chemical and process plants, and it can lead to domino effects due to thermal radiation. Emergency response is necessary to prevent fire escalation, in addition to safety barriers. In major industrial fire accidents, backup is usually required in emergency response, due to the lack of emergency response capacity of a single emergency response department. A method addressing the optimal allocation of emergency response forces, specifically considering the front-line departments, is developed in the present study. The emergency response of the front-line departments can reduce the thermal radiation received by the equipment adjacent to the primary fire and thus prolong the time to failure of the equipment, such that the backups have more time to get to the fire scene. The allocation analysis of emergency response teams is carried out considering the dynamic ttf of an adjacent equipment item resulting from the change in time of the thermal radiation received. A timed colored hybrid Petri-net (TCHPN) approach is proposed to model the emergency response process. The probability of preventing fire escalation is obtained from the TCHPN model and the optimal allocation of firefighting forces is determined. A case study illustrates the proposed approach, two scenarios are compared and results show that if a request for backup can be issued immediately according to the fire state, fewer emergency forces can be deployed on the front line department, such as the reduction from 6 emergency response teams to 3 teams to maintain the success probability of 0.91. The influence of other factors such as the position of fire departments and the layout of tanks on the allocation of emergency forces is also discussed.

Offshore Oil&Gas facilities are attractive targets of intentional malicious attacks (security attacks) that may trigger cascading events (e.g., the release and dispersion of hazardous material and/or energy, fires, explosions) with consequences on people, environment, and assets. The severity of these consequences is potentially similar to those arising from major accident scenarios originated by conventional safety-related causes. Current practice in managing the risk of security attacks mostly relies on qualitative or semi-quantitative procedures developed over the years in the offshore Oil&Gas industry. In the present study, a systematic quantitative procedure is developed, based on a Bayesian Network (BN) approach, for calculating the probability of success of physical security attacks, taking into account both preventive and mitigative security intervention strategies. The procedure addresses the specific framework of the offshore Oil&Gas industry. A case study concerning an offshore fixed Oil&Gas platform allowed us to demonstrate the quality of the results that can be achieved and their potential towards the improvement of the security of the installations considered.

Accidents triggered by natural events are becoming an increasing issue for policy-makers and industrial practitioners in the last decades. These events are called Natech (natural hazards triggering technological disasters), and their study is critical due to an increase in both the frequency and the magnitude of the consequences. Considering the possible failure of safety systems and the several equipment items potentially involved in the accidents due to the occurrence of multiple simultaneous failures and cascading events, the emergency response results to be a crucial aspect in the risk assessment and management of Natech events. Despite this, the emergency response can also be affected by the occurrence of natural events. The purpose of the present work is to understand the effects of natural events on emergency response. To reach this goal, some relevant past accidents are studied. The result of the analysis allows highlighting the effects of the natural events on the emergency response. Moreover, lessons learned on the critical elements of the emergency response are derived when considering Natech accidents and related cascading events.

Inherent safety design principles have been used extensively to design safer plants. However, to date their application is merely limited to technical changes in the process. An unexplored area of ISD is related to all the non-technical aspects linked to the safety of industrial plants, such as the Human and Organizational Factors, often referred to as HOFs. This study presents an Inherent Safety Decision Making framework (ISDDM), with the objective to provide a structured tool that can be easily used by industrial practitioners. The novelty of the ISDDM is the provision of a structured framework where purely technical and non-technical ISD alternatives are categorized and analyzed, to check if possible interactions exist among the alternatives proposed. To prove the usefulness of the model, a case-study addressing a carbon capture plant is carried out. This technology will play an important role in the reduction of CO₂ emission in carbon-intense industrial processes, implying a rapid scale-up of these plants in the next years, therefore justifying the analysis of this green technology. Adopting the proposed ISDDM, the improvements in terms of safety from the base case ranged from 58% to around 70% for the best design solution, while from an economic perspective, cost variations were comprised between 16% and 22%. The results of the study demonstrate that the framework can be rigorously implemented and that it provides a considerable number of additional alternatives due to the inclusion of HOFs in the analysis. In conclusion, an improved and adequate ISD is a key to environmental sustainability.

Cascading events in which an escalation triggered by a domino effect takes place are high-impact low-probability accidents that require specific prevention and mitigation in design and operation of industrial facilities and industrial parks. A domino effect phenomenon was responsible of several catastrophic accidents that affected the chemical and process industry, as well as critical infrastructures for energy and other industrial sectors where relevant quantities of hazardous substances are stored or processed. Although to date most of these events were caused by unintentional domino effects, related to process safety accidents, intentional, security-related, domino effects may cause similar scenarios. Significant regulatory requirements and technical standards address the prevention and mitigation of domino effects in the design and operation of industrial sites. However, the complexity of the topic and the limited data available result in a poor agreement on assessment procedures to address escalation hazards possibly resulting in domino scenarios. Even if methods, models, and tools developed now facilitate the quantitative assessment of domino scenarios in risk analysis and in the safety management of industrial sites, a number of open points still remain. In recent years, a constant progress was made in improving the quantitative methods for understanding and assessing domino effects, and in addressing and reducing the uncertainty. Recent research efforts dedicated to the exploration of domino effects allowed the development of methods for dynamic risk assessment. These may now be complemented by specific methods addressing the quantitative performance of safety barriers to prevent the escalation. In parallel, a relevant effort was dedicate to methods addressing the control and prevention of intentional domino effects. In the present chapter, an overview of the more active research areas in domino effects is provided, and the importance of innovation and new findings for the assessment and the management of the risk caused by cascading events triggered by domino effects is provided.

Accidents resulting in industrial fires in chemical and process installations and in industrial parks where relevant quantities of hazardous substances are stored or processed may cause domino effects. Probit models developed and used in a multitude of studies can provide the probability of equipment failure, but they do not consider the effect of multiple radiation sources, and thus fail to capture the effects of severe scenarios as those where multiple fires start at different times in different units. In the present study, a critical thermal dose for equipment failure is defined. A direct procedure for the calculation of ttf based on the critical thermal dose is then introduced, which is able to account for the time at which the different secondary fires start or are extinguished. This allows considering the effects of the primary and of several secondary fire scenarios in causing a domino effect, updating the time to failure on the basis of the dynamic evolution of multiple fire scenarios. The proposed approach is demonstrated through case-studies addressing fire-induced domino effects in an oil storage tank farm.

Industrial infrastructures, in particular those where hazardous substances are stored or handled, may be the target of malicious acts aiming at the disruption of normal operations. In the present study a toolbox of complementary and synergic techniques (Correspondence Analysis (CA), Fishbone Diagrams, Cause-Consequence Chains, Adversary Sequence Diagram, Root Cause Analysis) was applied to the in-depth analysis of physical security- and cybersecurity-related events that affected the process industry. The unprecedented original set of information obtained provides novel insights concerning these events. Clear correlations among security threats, including cyber-threats, and specific industrial sectors, as well as among the final scenarios and the different security threats from which they originate were identified by CA. In particular, vandalism resulted strongly correlated with the transportation of hazardous substances, and theft of materials with oil and gas pipelines. When considering chemical and petrochemical sites, cyber-attacks and the use of improvised explosives resulted to be the most common attack modes performed by the threat actors. Personnel and vehicle gateways resulted key elements when designing the Physical Protection System (PPS) of a facility. Insiders having the permission to enter the site bypass such controls, and were responsible of several successful attacks. Overall, the results confirm the concreteness of security-related events in the process industry and provide an original structured and detailed insight on the attack patterns experienced to date. Moreover, the results and the data obtained provide a novel set of baseline information for the application of SVA (Security Vulnerability Assessment) or SRA (Security Risk Assessment) methodologies in facilities where hazardous substances are stored or processed.

Industrial facilities where relevant quantities of hazardous chemicals are stored or processed may be direct or indirect targets of malicious acts of interference, including cyber or physical attacks. In the case of direct attacks, one or more process units may be damaged within the installation, often with the clear willingness of triggering domino effects escalating the primary event and possibly involving the entire installation. Also indirect attacks may be relevant, where the damage or destruction of neighboring sensible civilian targets is the primary aim of the action, but domino effects may be generated by the effects of the interaction with primary, large-scale accidents occurring outside the physical border of the facility. Escalation may result in cascading events that may heavily involve and damage the industrial facility of concern.

The authors regret that an imprecise statement was made in this article, and wish to offer this Corrigendum as clarification. The authors would like to apologise for any inconvenience caused. Imprecise statement: Misuri et al. (2018, pp. 70) state in their work: “Differently from EN [evidential network], CN [credal network] can be used both to conduct forward analysis and to update probabilities”. This statement implies that EN cannot be used for belief updating and should be mapped into a corresponding CN for that purpose. Correction: The foregoing statement may be correct for conventional ENs that are based on Dempster's combination rule, but does not hold true for the EN developed by Simon and co-workers (2008, 2009) based on Bayesian network (BN) inference algorithms (herein, BN-based EN). Simon and Weber (2009) explicitly mention in their work that the developed EN can be used for belief updating: “The computation mechanism is based on the Bayes theorem, which is extended to the representation of uncertain information according to the framework of Dempster-Shafer theory. Specific evidence (Hard evidence) is modeled by a mass of 1 on one of the focal elements of the frame of discernment. Non-specific evidence (Soft evidence) corresponds to a mass distribution on the focal elements of the frame of discernment.” Proof: Fig. 1 displays the BN-based EN developed in Misuri et al. (2018) for security vulnerability assessment of a process plant where the belief masses have been updated given the evidence “Attack = Success”. Misuri et al. (2018) used the BN-based EN for predicting the probability of a successful attack (forward analysis), but to update the probabilities (backward analysis) they mapped the BN-based EN into an equivalent CN and calculated the updated probabilities given “Attack = Success”. They used two packages, JavaBayes and GL2U, to implement the CN, concluding that JavaBayes results in more consisting updated probabilities with regard to the evidence (the 2nd column in Table 1). In the present corrigendum, we used the BN-based EN (Fig. 1) for belief updating given “Attack = Success”. The updated beliefs were subsequently used to calculate updated probability intervals (the 3rd column in Table 1), showing a good agreement between the results of CN and BN-based EN. Conclusion: The EN developed by Simon and co-workers (2008, 2009) based on BN can be used for belief mass updating the same way BN can be used for probability updating, with no need for using CN.

Security management of critical infrastructures is a complex task as a great variety of technical and socio-political information is needed to realistically predict the risk of intentional malevolent acts. In the present study, a methodology based on Limited Memory Influence Diagram (LIMID) has been developed for the protection of critical infrastructures via cost-effective allocation of security measures. LIMID is an extension of Bayesian network (BN) intended for decision-making, allowing for efficient modelling of complex systems while accounting for interdependencies and interaction of system components. The probability updating feature of BN has been used to investigate the effect of vulnerabilities on adversaries’ preferences when planning attacks. Moreover, the proposed methodology has been shown to be able to identify an optimal defensive strategy given an attack through maximizing defenders’ expected utility. Despite being demonstrated via a chemical facility, the methodology can easily be tailored to a wide variety of critical infrastructures.

Risk and Safety Management are crucial aspects in chemical industry and academic laboratories. From their rich experience in academic education and industrial practice, the authors present options for professional training addressing engineers and scientists at different career levels. The book informs about existing norms (OHSAS, ISO, etc.) and discusses examples from several countries. • Including sample material for illustrative courses and exams. • Presents discussion on how to implement risk management policies in education and research institutes.