Fire PSA Task 11: Detailed Fire Modelling Case Study

In 2016, Jacobsen Analytics Limited (JA) was contracted to support a project to perform fire PSAs for five units, across two nuclear power plants with an ambitious schedule. Working in an international team, JA's specific task was to provide expert support and to lead the Task 11 detailed fire modelling analyses, specifically for the Main Control Rooms (MCR), Main Relay Rooms and back-up Control and Main Relay Rooms. The analyses conducted previously for these rooms had yielded higher than expected risk-estimates when compared to similar power plants. Therefore, the task set for Jacobsen was to refine and update the analyses to obtain more realistic results.

A Fire PSA effectively overlays the potential likelihood and impact of fire events onto a plant logic model, which is built to represent how failures of safety equipment interact and impact safety functions, potentially leading to an inability to cool the reactor. The Fire PSA is comprised of a series of discrete and well defined tasks, which can be performed according to an acceptable methodology. For this particular project, the client wanted to follow the industry recognised and well documented NUREG/CR-6850 approach, which JA have extensive experience in the applying.

Fire PSA Task 11 in the NUREG/CR-6850 approach relates to detailed fire modelling, which is conducted for compartments that are potentially risk significant. The task considers the growth and propagation of a fire and the possibility of it being detected and suppressed before a specific target set is damaged. Three categories of fire scenarios are normally analysed, these being:

  1. general single compartment scenarios,
  2. MCR scenarios, and
  3. multicompartment scenarios.

The ultimate output of Task 11 is a set of fire scenarios, frequency of occurrence of those scenarios, and a list of target sets (in terms of fire PRA components) associated with the scenarios. For scenarios involving the MCR, the possibility of forced abandonment is also evaluated.

Initially, a thorough investigation of the previous analyses was conducted to determine the reason for the abnormally high risk estimates. It was concluded that a lack of detailed information that had led to a number of conservative assumptions and simplified modelling. The fire propagation analysis had not been fully being developed and conservative cable routing assumptions were being used.

Due to the challenging schedule and the development of 5 unit analyses in parallel, an approach was devised, for each room type, which incorporated the relevant methods from the latest FAQs, the aforementioned NUREG/CR-6850 and the more recently released NUREG-2178 guidance.

Generic fire scenario models representing the fire propagation of an ignition source plus any resultant secondary fires to certain target sets, were developed. The CFAST fire modelling code was used to determine the resultant environmental conditions for a range of different room dimensions, so as to determine the time to damage for each type of target and the time for operator abandonment.

Developing the analyses in a generic manner was essential, due to the large number of rooms and ignition sources that required evaluation, but it was also necessary due to way previous tasks in the Fire PRA had been conducted. One of the largest issues was missing, or incorrectly defined input information, an example of which was the location of ignition sources and their NUREG CR/6850 Task 6 classification; erroneous binning leads to a dilution of the fire ignition frequencies. While the ignition source information was easily amended, the lack of cable routing information, such as the location and the cable to component mapping was not easily determinable within the timeframe of the strict schedule. This ultimately meant that targets could not be correctly captured, which would likely lead to an underestimation of the risk.

Usually, target sets would be defined by a combination of analysing the geometric location of each target with respect to the ignition source's zone of influence and the hot gas layer temperature/elevation. With a lack of, or in some instances no cable routing data, it was extremely difficult to determine how complete this data was and it was therefore necessary to define conservative target sets to ensure that the resultant CDF was not being under-estimated. For the MCR, attempts were made to fill any gaps in the cable routing by using panel pictures and drawings to map components to panels. Where there were no pictures at all, the system associated with each panel was determined and failed. For the MRRs additional cable routing data has been added to complete the study. However, in some cases it was necessary to define entire banks of cabinets damaged, or even whole room damage for certain fire scenarios.

Further issues which required consideration included:

  • Concerns over the robustness of the internal events modelling
  • Treatment/consideration of spurious actuations possibly non-conservative.