Towards limiting potential domino effects from single flammable substance release in chemical complexes by risk-based shut down of critical nearby process units

Abstract

The explosion load is a significant escalation factor possibly influencing the potential occurrence of domino accidents in chemical plants. It is not economical to install explosion isolation systems (e.g., extinguishing barrier) for all equipment or process units across a chemical plant. Although shutting down all equipment or process unit can prevent an explosion, it may also cause further economic losses. To prevent domino accidents, the process unit that needs to be shut down accurately should be selected, and the normal operation of other units needs to be ensured. A method to select the process unit to be isolated based on the Dimensioning Accidental Load (DAL) is proposed. By calculating the occurrence probability and consequences of the accident scenarios, the DAL of the surrounding units is determined. DAL is used as the impact intensity of the accident unit on the surrounding units. The probit model is used to calculate the damage probability of surrounding units. The case analysis results show that the method of selecting the process unit to be isolated based on DAL quantifies the impact intensity of the exploded unit on surrounding units from probability and consequence. Under the premise of meeting the acceptable risk criteria, the method can determine which units should be shut down and which units can operate normally when a release accident occurs. While preventing domino accidents, economic losses caused by the shutdown of all process units are reduced and a theoretical basis for accident prevention and safe operation of the plant is provided.

Introduction

A chemical plant usually has an extensive storage and use of hazardous materials and a compact equipment layout. Once an accident occurs, it may cause severe casualties, property losses, and environmental damage (Vianello et al., 2019; Zhou and Reniers, 2018; Yang et al., 2015; Khan et al., 2016). When a flammable gas release accident occurs, an explosion accident may occur in the presence of the ignition source. The explosion at one process unit may severely impact the surrounding units and then escalate to a catastrophe through domino effects (Birk Michael, 2017). However, not all accidental releases of flammable gas and vapors create explosions. Most of the explosion accidents generate low or medium overpressure only (Chamberlian et al., 2019). When a release accident occurs in a plant, the shutdown of all process units are generally taken to prevent explosion accidents. However, shut down all process units may obviously cause significant economic losses. Therefore, the process unit that must be stopped and isolated during an accident should be reasonably determined to effectively terminate the accident propagation route and ensure the normal operation of other units.

It is a challenging task to select the process unit to be isolated to prevent domino accidents. The traditional method is generally based on experience, from the perspective of product process safety, determining which units should be shut down and isolated and which units can continue to operate according to the minimization of the influence between processes. However, the impact of specific physical location and explosion risk is not considered.

Domino accidents take place when an accident in a unit (primary unit) propagates to other units (secondary units) through the impact of escalation vectors (Kamil et al., 2019). Escalation vectors are physical effects such as overpressure in an explosion (Khakzad et al., 2013). Many scholars have studied domino accidents from the occurrence probability perspective and made outstanding contributions (Bagster and Pitblado, 1991; Necci et al., 2015; Khan and Abbasi, 1998; Cozzani and Salzano, 2004). Cozzani et al. (2005) developed a systematic procedure for the quantitative assessment of domino effect risks. Reniers et al. (2009) used a game-theoretic approach to interpret and model behavior of chemical plants within chemical clusters while negotiating and deciding on domino effects prevention investments. A quantitative risk assessment method for chemical domino accidents based on field theory and Monte Carlo simulation is proposed by He and Weng (2020). This method aims to obtain the dynamic distribution of individual risk caused by the domino accident. Zeng et al. (2020) developed a new method to provide more accurate probabilities related to domino effects, by considering the temporal evolution of escalation vectors caused by time-dependent factors. Naderpour and Khakzad (2018) proposed a Natech (natural hazard triggering technological disasters) risk assessment methodology that relies upon Bayesian network capabilities and takes into account the potential Natech domino effects. Ding et al. (2020) proposed a novel approach to model the spatial-temporal evolution and performed a risk analysis of fire-induced domino effects based on synergistic effect and accident evidence. These studies focus on finding out the development path of Domino accidents and probability, which lays a foundation for accident prevention and control and emergency decision-making in a chemical plant. However, in consequence analysis, only the consequences of accidents under the worst conditions are generally analyzed, leading to excessive risk assessment (Paik et al., 2011). From the perspective of the accident consequences and corresponding probability, this study investigates how to select the process unit to be isolated to prevent a domino accident based on its predicted probability and consequence.

Risk is determined by the probability and the consequence of an accident. Only considering the consequence of the accident may lead to excessive measures, such as the shutdown of all process units. When the explosion load and probability of a unit affected by the initial unit exceed the threshold, the unit is selected as an process unit to be isolated. NORSOK Z-013 (2010) provides a framework for explosion risk analysis but does not provide specific analysis methods and cases. There are many methods to assess the impact of an explosion, such as the Multi-energy method, the TNT equivalent method, the Baker-Strehlow method, and CFD simulations. (van den Berg and Lannoy, 1993; Qiao and Zhang, 2010; Qi et al., 2019; Chen et al., 2020; Zhang et al., 2020; Horvat, 2018; Moen et al., 2019). The multi-energy method is widely used in the two-dimensional model, which comprehensively considers the turbulence acceleration and gas activity. However, it is difficult for the two-dimensional model to consider the impact of the complex equipment and piping system of the plant on the explosion overpressure. In recent years, computational fluid dynamics (CFD) software has developed rapidly, such as AutoReaGas, FLUENT and FLACS. Since the CFD model can well represent the real physical environment, it has increasingly become the mainstream of explosion risk analysis (Huang et al., 2019; Shen et al., 2020). Hansen et al. (2016) used FLACS to evaluate the explosion overpressure of equipment, pipelines, and critical buildings and convert the overpressure into actual forces, guiding the design strength of the equipment. Baalisampang et al. (2019) used FLACS to analyze the integrated consequence of hydrocarbon release, fire, explosion, and dispersion of combustion products. Rui et al. (2021) investigated the effect of low vent burst pressure on the overpressure buildup and flame evolution during the vented methane-air deflagrations in a 1 m³ rectangular vessel, and used FLACS to validated the experimental data. Li and Hao (2018) proposed a combined CFD approach on far-field pressure prediction based on experiments and FLACS. Dadashzadeh et al. (2014) proposed a new method to quantify the risk of combustion products dispersion phenomenon in a confined or semi-confined facility. Li and Hao (2019) investigated the explosion pressure and impulse from large-scale explosions by using experiments and FLACS.

The explosion load is affected by the gas cloud volume and position, and ignition position. The volume and position of the gas cloud are affected by factors such as wind speed, wind direction, leak rate, release location, release direction, etc. The combination of these factors results in a large number of accident scenarios. Although the CFD model can represent the real physical environment well, the existing study is limited to the worst-case simulation or some selected hypothetical scenarios for analysis due to the computer resources and time limitations. This increases the uncertainty of the scenario selection and thus makes the results not representative.

The present study aims to propose a method to a) select representative scenarios and reduce the uncertainty of scenario selection, and b) to comprehensively analyze the risk of the affected units from the perspectives of accident consequence and probability. The dimensioning accidental load (DAL) of each unit under the influence of accident units is calculated, and DAL is used as the input of the probit model to calculate the potential damage probability of each affected unit. When the risk is high, the unit is selected as an process unit to be isolated.

The remaining parts of this paper are organized as follows. The isolation idea in the process industry is presented in Section 2. A brief description of the proposed method, including scenario development, explosion risk analysis, and how to select process unit to be isolated, is shown in Section 3. The case study is presented in Section 4. The assumptions and limitations are discussed in Section 5. Finally, conclusions are drawn in Section 6.