A lot of companies that operate gas-fired boilers are in the process of considering and installing gas detection systems to protect their plants. In the United States, and many other countries, the design of boilers is in accordance with National Fire Protection Agency (NFPA) Standard 85 – Boiler and Combustion Systems Hazards Code. This standard covers a comprehensive set of topics related to boiler system design, and is widely accepted and adopted. This code recommends gas detection, but does not provide a lot of insight into the design of the detection system. The standard basically states, “Consideration should be given to detecting and monitoring combustible gases in areas where they are likely to accumulate.” In addition, this standard also refers the reader to NFPA 54 – The National Fuel Gas Code, for more information on the fuel gas supply system design. NFPA provides a bit more information, but not a substantial design technique. In essence, the guidance for gas leak detection is as follows, “184.108.40.206 the leakage shall be located by means of an approved gas detector, a noncorrosive leak detection fluid, or other approved leak detection methods. Matches, candles, open flames, or other methods that provide a source of ignition shall not be used.
Industry has noted gas leaks in boilers to be a significant safety hazard, and in many cases have implemented corporate standards that require combustible gas detection in the vicinity of boilers. Furthermore, there is often a requirement that a confirmed gas leak should result in automatic isolation of the fuel gas supply. But, as with the NFPA standards for fuel gas system design, no concrete guidance on detector placement is given. As is the state of the art of the recent past, most of this detector placement has been performed by rules of thumb and expert judgment. This resulted in inconsistent designs with variable numbers of detectors for identical installations across an organization. Industry can do better. By using performance based detection methods, designs that ensure tolerable risk by quantitatively verifying the degree of coverage of the gas detection system can be employed by industry, reducing risk and gas detection system lifecycle costs.
Kenexis recommends use of the performance-based approaches identified in ISA Technical Report Tr84.00.07, where coverage targets are determined and quantitatively verified. The specific methods for each type of boiler system will depend on its size, use, and location. It is expected that boilers that are installed entirely outdoors will require less gas detection, if any, due to the ability for gas leaks to be dispersed to the atmosphere without accumulating to a concentration beyond the Lower Flammability Limit (LFL). Boilers that are located inside closed buildings pose the highest risk because it is easier for fuel gas to accumulate in a closed space, and the degree of confinement will also contribute to the propensity for an ignition to cause a devastating vapor cloud explosion as opposed to the less severe flash fire that would be expected out of doors.
In closed buildings the dispersion of a gas from a leak (especially a small one) will be dominated by the effect of the building’s HVAC system. Traditional “similarity” dispersion modeling will be entirely useless in this case. In order to understand the size and location of a gas cloud resulting from design basis leaks of a boiler fuel supply system in an indoor location, a Computational Fluid Dynamics (CFD) style dispersion analysis is required. In order to perform detector placement, an engineering team would model the physical characteristics of the room in which the boiler is located, including the ventilation locations and rates. Then a series of design basis releases, at several representative locations would be modeled.
For this blog, my colleague Sean Cunningham has prepared a sample study. The first step is to define the facility. This is done by 3D model creation, or import from a 3D CAD program – along with definition of the extents of the room and the HVAC definition. The result of this task is shown in the following figure which shows a 3D model of a single boiler room with an HVAC inlet and outlet.
After the model is built, the leak scenarios need to be defined and input to the CFD modelling software. The following figure shows details of the leak.
After the room and leak source are defined, the CFD model is executed. The CFD modelling software generates a time series of gas concentrations at a matrix of three dimension points in the space of the model. This time series can then be used to created graphs of concentration versus time at various points, or even create 3D video of concentration versus time as a 3D isosurface or a 2D slice showing concentration isopleths. The figure below shows the isosurface made by the lower flammability limit (LFL) for an instant in time.
The next figure shows the concentration profile for a single elevation. This figure shows colors as concentration isopleths (lines of constant concentration). The figure is shown near the ceiling. The elevation was selected because natural gas is buoyant, and expected to rise after a release, making a higher elevation a better choice for detection location.
All of the figures that are shown above are for a single instant in time, but as a release occurs the concentrations vary. In order to get a better appreciation of the release characteristics, a video of the modeled release can be created and viewed. The video below contains both the LFL isosurface, and the ceiling elevation concentration profile.
During the course of modeling, several potential detector locations can be defined. Based on the results of the modeling, for each design basis release that is modeled, the concentration of gas at the detector can be graphed for the release duration. The objective of the detector placement would be to select the minimal set of detectors that would ensure that 100% of the desig
n basis scenarios are detected. In doing so, the design team can be comfortable that the number of detectors and their selected location are appropriate for all foreseeable release scenarios. This approach will optimize the detector placement, ensuring effective coverage and minimizing the number of detectors that are essential to safety. Furthermore, this type of modeling can be done quite cost effectively, especially if many boiler rooms with similar equipment need to be modeled.