Probabilistic model data of time-dependent accident scenarios for a mixing tank mechanical system

This article presents the risk assessment of a mixing tank mechanical system based on the failure probabilities of the components. Possible component failures can cause accidents which evolve over multiple time stages and can lead to system failure. The consequences of these accident scenarios are analyzed by quantifying the failure probabilities and severity of their outcomes. Illustrative costs and updated failure probabilities are provided to evaluate preventive safety measures. Data refers to the results of the Bayesian model presented in our research article (Mancuso et al., 2019).


Data
This article presents the probabilistic model data of the time-dependent accident scenarios for a mixing tank mechanical system. Specifically, we revisit the earlier analyses of the accident scenarios by Khakzad et al. [2] to illustrate the methodology presented in our research article [1]. One of such accident scenarios occurred on 14 June 2006 at Universal Form Clamp in Bellwood (Illinois, U.S.) through a vapor cloud ignition [3]. Table 1 shows the failure probabilities of Alarm and Sprinkler for different ways of activating such components during an accident. In particular, the activation occurs if the vapor is ignited or if there is a specific amount of vapor concentration in the air, even though the vapor is not ignited.
Based on the analyses by Khakzad et al. [2], Table 2 lists the system components and their failure probabilities. In addition, we assume that the activation of Sprinkler reduces the probability of delayed ignitions by 50%, as detailed in Table 3 (last row, first and second columns). For this reason, the activation of the Sprinkler for a vapor concentration in the air could prevent delayed ignitions. Table 4 lists the nine possible outcomes of the accident scenarios where the state Safe represents the outcome following the non-occurrence of the system failure (Vapor ¼ Controlled). The other outcomes are caused by malfunctions of some system components. Due to the activation of Sprinkler, accident consequences C 1 and C 2 are less severe than C 3 and C 4 , respectively. This information is helpful in eliciting the disutility functions to specify the ranking of the outcome severity. The last column of Table  4 shows illustrative disutility values that quantify the severity of the outcomes.
Based on the failure probabilities in Table 2, the Bayesian model computes the occurrence probabilities of the outcomes of the accident scenarios, reported in Table 5 for each time stage. The deployment of preventive safety measures on some selected components mitigates the risk of the negative outcomes. Table 6 lists the alternative preventive safety measures (second column) that affect the occurrence of failures of specific components (first column). The last two columns of Table 6 report illustrative costs and updated failure probabilities of the components. In particular, the preventive safety measure Synergy refers to a combination of Calibration test and Sensor: if both Specifications Value of the data The failure probabilities of the components of a mixing tank mechanical system can be used for benchmarking in future research.
Examples of conditional probability tables illustrate the modelling of time-dependent accident scenarios. Novel applications for probabilistic risk assessment are possible based on the data in this article.  Table 3 Conditional probabilities of Ignition at t > 0 (t refers to the time stage of the Bayesian model).  measures are installed, this synergy effect yields more benefits than installing independent measures. The updated failure probabilities of Sprinkler and Alarm refer to the two different failure scenarios detailed in Table 1.

Experimental design, materials, and methods
The failure probabilities of the components in Table 2 are provided by the article by Khakzad et al. [2]. Gates represents logic structures of the Bayesian model in our research article [1]. The failure probabilities in Table 6 have been obtained by reducing the initial failure probability of the components, based on a specific reduction rate for each preventive safety measure. These values illustrate the viability of the Bayesian model [1], but do not represent any actual system. The occurrence probabilities of the outcomes of the accident scenarios have been computed by GeNIe Modeler [4] through the Dynamic Bayesian Network presented in our research article [1]. Finally, the severity of the outcomes has been quantified through the trade-off weighing approach SWING [5]. Table 5 Probabilities of accident outcomes at each time stage (C refers to the accident consequences).